CN114332914A - Personnel feature identification method, device and computer-readable storage medium - Google Patents
Personnel feature identification method, device and computer-readable storage medium Download PDFInfo
- Publication number
- CN114332914A CN114332914A CN202111435354.5A CN202111435354A CN114332914A CN 114332914 A CN114332914 A CN 114332914A CN 202111435354 A CN202111435354 A CN 202111435354A CN 114332914 A CN114332914 A CN 114332914A
- Authority
- CN
- China
- Prior art keywords
- neural network
- network model
- person
- personnel
- multitask
- 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
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000003062 neural network model Methods 0.000 claims abstract description 96
- 238000012549 training Methods 0.000 claims abstract description 57
- 238000001514 detection method Methods 0.000 claims abstract description 31
- 238000010276 construction Methods 0.000 claims abstract description 11
- 238000013528 artificial neural network Methods 0.000 claims description 18
- 238000005457 optimization Methods 0.000 claims description 8
- 238000011478 gradient descent method Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 description 22
- 238000010586 diagram Methods 0.000 description 11
- 230000006399 behavior Effects 0.000 description 7
- 230000000694 effects Effects 0.000 description 7
- 238000013135 deep learning Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000012804 iterative process Methods 0.000 description 4
- 238000013473 artificial intelligence Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 206010000117 Abnormal behaviour Diseases 0.000 description 1
- 230000037237 body shape Effects 0.000 description 1
- 210000000481 breast Anatomy 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000002604 ultrasonography Methods 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention provides a personnel feature identification method, a device and a computer readable storage medium, wherein the personnel feature identification method comprises the following steps: s100: constructing a personnel multitask neural network model with personnel key point detection; s200: inputting and training a person multitask neural network model by taking a data set as a training sample; s300: randomly optimizing a loss function on training data in each iteration of the training personnel multitask neural network model, and finishing the training of the personnel multitask neural network model after the iteration is carried out for a preset number of times; s400: acquiring personnel characteristics through the trained personnel multitask neural network model and identifying multidimensional characteristics; s500: and (5) collecting the multi-dimensional features to construct a person integral portrait. According to the personnel feature identification method provided by the invention, the multitask neural network model is optimized, so that the construction efficiency and the accuracy of the personnel integral portrait are improved.
Description
Technical Field
The present invention relates to the field of artificial intelligence technology, and in particular, to a method and an apparatus for identifying a person feature, and a computer-readable storage medium.
Background
In recent years, with the development of technologies for identifying and detecting pedestrians in videos and images by using an artificial intelligence technology, the development of intelligent video monitoring processing is greatly promoted, the main characteristic information of pedestrians mainly comprises human faces, human bodies and skeleton information, wherein the obtained human faces can be used as key identification information for identifying the identities of the pedestrians and the like; the method comprises the following steps of obtaining human body attribute information, obtaining appearance information of a human body, and performing functions of searching images by images, tracking personnel trajectories and the like; the pedestrian skeleton information is acquired, the pedestrian skeleton information can be used for detecting various behaviors of people, such as abnormal behaviors of fighting, carrying, waving hands for help and the like, the human face, the human body attribute and the skeleton can carry out all-dimensional feature description on the pedestrian, and the pedestrian skeleton information is the most important three parts in the pedestrian monitoring information.
In the prior art of identifying and detecting people from multiple dimensions based on an artificial intelligence technology, three detection tasks of a human face, a human body and a skeleton are trained by different network models, the three detection tasks are single in function and long in training time consumption, and the generalization capability of a system is often deteriorated due to insufficient sample capacity, so that the detection efficiency is low, the identification speed is low, the three identification tasks can be mutually supported, and the relationship among the three tasks is not effectively utilized.
Disclosure of Invention
The invention aims to solve the technical problems of how to improve the efficiency in a multitask neural network and how to identify the multidimensional characteristics of personnel and integrally form a portrait by a multitask neural network model, and provides a personnel characteristic identification method, a personnel characteristic identification device and a computer readable storage medium.
The personnel feature identification method provided by the invention comprises the following steps:
constructing a personnel multitask neural network model with personnel key point detection;
inputting and training a person multitask neural network model by taking a data set as a training sample;
randomly optimizing a loss function on training data in each iteration of the training personnel multitask neural network model, and finishing the training of the personnel multitask neural network model after the iteration is carried out for a preset number of times;
acquiring personnel characteristics through the trained personnel multitask neural network model and identifying multidimensional characteristics;
and (5) collecting the multi-dimensional features to construct a person integral portrait.
According to the personnel feature identification method provided by the invention, the loss function on one piece of training data is randomly optimized in the iterative process of the multitask neural network model, so that the parameter updating speed of each iteration is greatly accelerated, the training of the multitask neural network model can be completed more efficiently, and further the personnel features are obtained and the multidimensional features are identified through the trained personnel multitask neural network model, so that the construction efficiency and the accuracy of the personnel integral portrait are improved.
According to some embodiments of the invention, the human multitasking neural network model is a parametric hard-shared multitasking neural network model.
In some embodiments of the invention, the human multitasking neural network model employs a HRNet-like neural network structure.
According to some embodiments of the invention, the human multitasking neural network model comprises FeatureNet, BodyNet, FaceHeadNet, HandHeadNet, wherein FeatureNet is a low-level shared feature extracted by BodyNet, FaceHeadNet, HandHeadNet.
In some embodiments of the present invention, the randomly optimizing a loss function on a piece of training data in each iteration of the training person multitask neural network model includes:
adding a multitask loss layer into the staff multitask neural network model;
in the multitask loss layer, regularization constraint is carried out on each task loss in the staff multitask neural network model, and the total loss of the multitask loss layer is calculated by combining noise and each task loss;
and solving the multitask neural network model by using a random gradient descent method to obtain an optimized result.
According to some embodiments of the invention, the human multitasking neural network model comprises: the method comprises the steps that a person key point detection multitask neural network model and a person key point application multitask neural network model are adopted;
the multi-dimensional features include: basic characteristics of the person to be detected and behavior characteristics of the person to be detected;
the method for acquiring the multi-dimensional characteristics of the personnel characteristic recognition through the trained personnel multitask neural network model specifically comprises the following steps:
the person key point detection multitask neural network model detects the basic characteristics of the person to be detected,
the personnel key points apply a multitask neural network model to detect the behavior characteristics of the personnel to be detected,
the basic features comprise key points of the body, the face and the limbs of the person to be detected, and the behavior features comprise posture estimation of the person to be detected.
The present invention provides a computer-readable storage medium storing computer instructions for performing a person feature identification method as in some embodiments of the present invention when executed by a computer.
The invention provides a personnel feature recognition device, comprising:
the building module is used for building a person multitask neural network model with person key point detection;
the input module is used for inputting and training the staff multitask neural network model by taking a data set as a training sample;
the optimization module is used for randomly optimizing a loss function on training data in each iteration of training the staff multitask neural network model and finishing the training of the staff multitask neural network model after the preset number of iterations;
the identification module is used for acquiring personnel characteristics through the trained personnel multitask neural network model and identifying multidimensional characteristics;
and the construction module is used for collecting the multi-dimensional features and constructing the person integral portrait.
According to the personnel feature recognition device provided by the invention, the loss function on one piece of training data is randomly optimized in the iterative process of the multitask neural network model, so that the parameter updating speed of each iteration is greatly accelerated, the training of the multitask neural network model can be completed more efficiently, and further the personnel features are obtained and the multidimensional features are recognized through the trained personnel multitask neural network model, so that the construction efficiency and the accuracy of the personnel integral portrait are improved.
In some embodiments of the present invention, the person feature identification apparatus further includes:
a sensor providing the person characteristic;
the control module acquires the personnel integral image and performs early warning;
and the studying and judging module is used for receiving the early warning information sent by the control module.
According to some embodiments of the invention, the human multitasking neural network model is a parameter hard-shared HRNet-like neural network structure.
Drawings
FIG. 1 is a schematic flow chart of a person feature identification method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram illustrating a partial flow chart of a person feature identification method according to an embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating a partial flow chart of a person feature identification method according to an embodiment of the present invention;
FIG. 4 is a schematic diagram illustrating a flow of detecting a key point of a person according to an embodiment of the present invention;
FIG. 5 is a schematic diagram illustrating a flow of applying a key point to a person according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of a deep learning network based on hard constraint multitask learning;
FIG. 7 is a schematic diagram of a deep learning network based on soft constraint multitask learning;
FIG. 8 is a schematic flow chart of a multitask neural network personnel feature detection model of a personnel feature identification method according to an embodiment of the invention;
fig. 9 is a schematic flow chart of a HRNet-like basic network structure in a person feature identification method according to an embodiment of the present invention;
FIG. 10 is a diagram illustrating a multitasking loss layer of a method for identifying characteristics of a person according to an embodiment of the present invention;
FIG. 11 is a schematic diagram of key point detection in a person feature identification method according to an embodiment of the present invention;
FIG. 12 is a schematic diagram of a person key point of a person feature identification method according to an embodiment of the present invention;
FIG. 13 is a schematic structural diagram of a person feature recognition apparatus according to an embodiment of the present invention;
fig. 14 is a schematic structural diagram of a person feature identification device according to an embodiment of the present invention.
Detailed Description
To further explain the technical means and effects of the present invention adopted to achieve the intended purpose, the present invention will be described in detail with reference to the accompanying drawings and preferred embodiments.
The invention solves the problem of the traditional single task method by applying a multitask neural network, the multitask neural network belongs to multitask learning, the multitask learning is a machine learning method which improves the effect of each task by combining data of a plurality of related tasks, and the multitask learning is defined as follows: given n learning tasks, of which two or more may be relevant, the goal of multi-task learning is to extract relevant information from all tasks to improve the effectiveness of one or more of the tasks in the model. The core idea of the multi-task learning is to combine a plurality of tasks to share common information of the tasks, especially under the condition that training samples are insufficient, the multi-task learning can fully exert the advantages of the multi-task learning and mine information as much as possible, so that the invention utilizes the characteristic of the multi-task learning to improve the effect of deep learning network under the condition that breast ultrasound images are few.
In order to solve at least some technical problems of how to improve the efficiency in the multitask neural network and how to identify the multidimensional characteristics of the person and construct the integral portrait by the multitask neural network model, as shown in fig. 1, the invention provides a person characteristic identification method, a device and a computer readable storage medium.
The personnel feature identification method provided by the invention comprises the following steps:
s100: and constructing a person multitask neural network model with person key point detection.
S200: and inputting and training a human multitask neural network model by taking the data set as a training sample.
S300: in each iteration of the training personnel multitask neural network model, a loss function on training data is randomly optimized, and training of the personnel multitask neural network model is completed after iteration for a preset number of times.
S400: and acquiring personnel characteristics through the trained personnel multitask neural network model and identifying multidimensional characteristics.
S500: and (5) collecting the multi-dimensional features to construct a person integral portrait.
According to the personnel feature identification method provided by the invention, the loss function on one piece of training data is randomly optimized in the iterative process of the multitask neural network model, so that the parameter updating speed of each iteration is greatly accelerated, the training of the multitask neural network model can be completed more efficiently, and further the personnel features are obtained and the multidimensional features are identified through the trained personnel multitask neural network model, so that the construction efficiency and the accuracy of the personnel integral portrait are improved.
According to some embodiments of the invention, the human multitasking neural network model is a parametric hard-shared multitasking neural network model.
Specifically, as shown in fig. 6 and 7, there are two sharing ways of the multi-task learning in the deep learning network, namely, a hard constraint multi-task learning method and a soft constraint multi-task learning method, where the hard constraint is suitable for a case where the association between several tasks is large, and the soft constraint can be used when the association between the tasks is not large because the constraint on the parameters is small. The hard constraint network requires that different tasks completely share parameters at the bottom layer and use the same characteristics, and each task uses own characteristics and parameters at the high layer, which is the most common mode for multi-task learning in the neural network.
The soft constraint method does not require the bottom-layer characteristics and parameters of a plurality of tasks to be completely the same, but regularizes shared bottom-layer parameters to ensure the similarity of the parameters.
The invention aims to improve the pedestrian feature detection effect through a multitask method, and the computer mainly refers to the characteristics of human body shape, human face and the like when automatically detecting the pedestrian features, so that the pedestrian skeleton detection, the human face detection and the human body detection have strong correlation, and a multitask learning method based on hard constraint is used when designing a multitask model.
In some embodiments of the invention, the human multitasking neural network model employs a HRNet-like neural network structure.
According to some embodiments of the invention, the human multitasking neural network model comprises FeatureNet, BodyNet, FaceHeadNet, HandHeadNet, wherein FeatureNet is a low-level shared feature extracted by BodyNet, FaceHeadNet, HandHeadNet.
Specifically, the invention designs a trigeminal scaling neural network (3T-zoomnet) structure based on a parameter hard sharing mechanism, as shown in FIG. 8, the trigeminal scaling neural network is a single network model and consists of FeatureNet, BodyNet, FaceHeadNet and HandHeadNet. FeatureNet extracts low-level shared features for BodyNet, faceHeadNet, HandHeadNet. BodyNet predicts body/foot keypoints and provides approximate regions of faces and hands for faceHeadNet and HandHeadNet, which detect face and hand keypoints, respectively, based on higher resolution features. The 3T-zoomnet is applied to all hidden layers of all tasks through a hard sharing mechanism of parameters, while retaining the task-related output layer reduces the risk of overfitting. The more tasks learned at the same time, the more tasks the same representation can be captured by the model, resulting in less risk of overfitting on the original task.
In the trigeminal scaling neural network (3T-zooman) structure, as shown in fig. 9 and fig. 10, FeatureNet, body net, facehead net, and hand headnet adopt the HRNet-like neural network structure, the HRNet has a parallel structure, and can maintain the high-resolution representation at any time, and the HRNet does not only rely on recovering the high-resolution representation from the low-resolution representation. The HRNet can obviously improve the gesture recognition effect, and has better effects in three tasks of key point detection, gesture estimation and multi-person gesture estimation of a COCO data set.
In some embodiments of the present invention, as shown in fig. 2, in step S300, in each iteration of training the human multitask neural network model, a loss function on a piece of training data is randomly optimized, including:
s310: and adding a multitask loss layer into the human multitask neural network model.
S320: and in the multitask loss layer, regularization constraint is carried out on each task loss in the human multitask neural network model, and the noise and each task loss are combined to calculate the total loss of the multitask loss layer.
S330: and solving the multitask neural network model by using a random gradient descent method to obtain an optimized result.
Specifically, in order to solve the technical problem that training time is long due to the fact that problems exist in optimization of a loss function in the prior art, a multitask neural network loss function mathematical model is built according to requirements of multitask multidimensional personnel feature identification, in the multitask detection loss function of personnel feature detection, in order to avoid the defects of a traditional method, namely multitask loss is simply weighted and summed on each task, the weight of each task is equal or manually set, some tasks easily occupy the gradient descending leading position, other tasks cannot be fully optimized, the optimization effect of multitask learning is greatly reduced, task noise is introduced from the perspective of task-dependent uncertainty, and on the basis of a Gaussian process, regularization constraint is conducted on loss (L W) of each task.
Wherein, the improvement loss function of the single task is as follows:
in step S310, weighting is performed
And a penalty item log (sigma) is studied, a multi-task loss layer is designed, three detection tasks of face key points, human body key points and skeleton key points of a person to be detected are taken as examples, the output of the three tasks is taken as the input of the loss layer, and the output of the three tasks is taken as the input of the loss layer
And log (σ) as a network parameter. The loss of the three tasks is respectively Lface、Lbody、LskelIntroducing noises 1, 2, 3, and calculating the total loss of the loss layer according to the following formulaCalculating:
in step S320, the multi-task neural network model is solved to obtain an optimization result of the variable to be optimized, including performing optimization solution by using a Stochastic gradient descent (Stochastic gradient parameter) method, where the Stochastic gradient descent method is different from batch gradient descent, and the Stochastic gradient descent is to update the network parameters by using one sample for each iteration. So that the training speed is increased. The objective function for one sample is:
and (3) solving a partial derivative of the objective function:
updating network parameters:
because the loss function on one training data is randomly optimized in each iteration instead of the loss function on all the training data, the updating speed of each parameter is greatly accelerated, and the detection result of the key points of the multitask personnel is as shown in fig. 8. And (4) according to the multitask neural network output, making a data set, wherein the data set comprises a data set which is made and disclosed by the data set, and the data set can comprise COCO, COCO-wholebody, MPII-TRB, OCHuman and the like.
According to some embodiments of the invention, the human multitasking neural network model comprises: the personnel key point detection multitask neural network model and the personnel key point application multitask neural network model, and the multidimensional characteristics comprise: the basic characteristics of the personnel to be detected and the behavior characteristics of the personnel to be detected.
As shown in fig. 3 to 5, 11 and 12, in step S400, the trained staff multitask neural network model is used to obtain multidimensional features of staff feature recognition, specifically:
s410: the person key point detection multitask neural network model detects basic characteristics of a person to be detected.
S420: and the personnel key points apply a multitask neural network model to detect the behavior characteristics of the personnel to be detected.
S430: the basic characteristics comprise key points of the body, the face and the limbs of the person to be detected, and the behavior characteristics comprise posture estimation of the person to be detected.
The present invention provides a computer-readable storage medium storing computer instructions for performing a person feature identification method as in some embodiments of the invention when executed by a computer.
The present invention provides a person feature recognition device, as shown in fig. 13, including: the system comprises a construction module, an input module, an optimization module, an identification module and a construction module, wherein the construction module is used for constructing a personnel multitask neural network model with personnel key point detection, the input module is used for inputting and training the personnel multitask neural network model by taking a data set as a training sample, the optimization module is used for randomly optimizing a loss function on a piece of training data in each iteration of the training personnel multitask neural network model, training of the personnel multitask neural network model is completed after iteration for a preset number of times, the identification module is used for acquiring personnel characteristics and identifying the multidimensional characteristics through the trained personnel multitask neural network model, and the construction module is used for collecting the multidimensional characteristics and constructing an integral personnel portrait.
According to the personnel feature recognition device provided by the invention, the loss function on one piece of training data is randomly optimized in the iterative process of the multitask neural network model, so that the parameter updating speed of each iteration is greatly accelerated, the training of the multitask neural network model can be completed more efficiently, and further the personnel features are obtained and the multidimensional features are recognized through the trained personnel multitask neural network model, so that the construction efficiency and the accuracy of the personnel integral portrait are improved.
In some embodiments of the present invention, as shown in fig. 14, the person feature identification apparatus further includes: the system comprises a sensor, a control module and a study and judgment module, wherein the sensor provides personnel characteristics, the control module acquires personnel integral portrait and performs early warning, and the study and judgment module is used for receiving early warning information sent by the control module.
According to some embodiments of the invention, the human multitasking neural network model is a parameter hard-shared HRNet-like neural network structure.
In summary, in the research of face, body and skeleton detection in the prior art, three detection tasks are trained by using different network models, the function is single, the training time is long, and the generalization capability of the system is often deteriorated due to insufficient sample capacity, so that the detection efficiency is low, the recognition speed is slow, the three recognition tasks can be mutually supported, and the relationship among the three tasks is not effectively utilized.
The invention provides a method and a device for identifying characteristics of multi-task and multi-dimensional personnel, wherein compared with the traditional method, the method for identifying the multi-task portrait improves generalization performance by utilizing the correlation among tasks, shares characteristic information, balances noise difference and restrains each other, reduces the risk of overfitting of a single task, finally improves the performance of all or part of learning tasks, and effectively solves the problems of the existing method, such as single function, long training time consumption, poor generalization capability caused by insufficient samples, low detection efficiency, low identification speed and underutilization of the commonality relation among the three tasks.
While the invention has been described in connection with specific embodiments thereof, it is to be understood that it is intended by the appended drawings and description that the invention may be embodied in other specific forms without departing from the spirit or scope of the invention.
Claims (10)
1. A person feature identification method is characterized by comprising the following steps:
constructing a personnel multitask neural network model with personnel key point detection;
inputting and training a person multitask neural network model by taking a data set as a training sample;
randomly optimizing a loss function on training data in each iteration of the training personnel multitask neural network model, and finishing the training of the personnel multitask neural network model after the iteration is carried out for a preset number of times;
acquiring personnel characteristics through the trained personnel multitask neural network model and identifying multidimensional characteristics;
and (5) collecting the multi-dimensional features to construct a person integral portrait.
2. The person feature identification method according to claim 1, wherein the person multitasking neural network model is a parametric hard-shared multitasking neural network model.
3. The person feature identification method according to claim 1, wherein the person multitasking neural network model adopts a HRNet-like neural network structure.
4. The person feature identification method according to claim 3, wherein the person multitask neural network model comprises FeatureNet, BodyNet, FaceHeadNet, and HandHeadNet, wherein the FeatureNet is a low-level shared feature extracted by BodyNet, FaceHeadNet, and HandHeadNet.
5. The method for identifying human features according to claim 1, wherein the randomly optimizing a loss function on a piece of training data in each iteration of training the human multitasking neural network model comprises:
adding a multitask loss layer into the staff multitask neural network model;
in the multitask loss layer, regularization constraint is carried out on each task loss in the staff multitask neural network model, and the total loss of the multitask loss layer is calculated by combining noise and each task loss;
and solving the multitask neural network model by using a random gradient descent method to obtain an optimized result.
6. The person feature identification method according to claim 1, wherein the person multitasking neural network model comprises: the method comprises the steps that a person key point detection multitask neural network model and a person key point application multitask neural network model are adopted;
the multi-dimensional features include: basic characteristics of the person to be detected and behavior characteristics of the person to be detected;
the method for acquiring the multi-dimensional characteristics of the personnel characteristic recognition through the trained personnel multitask neural network model specifically comprises the following steps:
the person key point detection multitask neural network model detects the basic characteristics of the person to be detected,
the personnel key points apply a multitask neural network model to detect the behavior characteristics of the personnel to be detected,
the basic features comprise key points of the body, the face and the limbs of the person to be detected, and the behavior features comprise posture estimation of the person to be detected.
7. A computer-readable storage medium storing computer instructions for performing the person feature identification method according to any one of claims 1 to 6 when the computer instructions are executed by a computer.
8. A person feature identification device, comprising:
the building module is used for building a person multitask neural network model with person key point detection;
the input module is used for inputting and training the staff multitask neural network model by taking a data set as a training sample;
the optimization module is used for randomly optimizing a loss function on training data in each iteration of training the staff multitask neural network model and finishing the training of the staff multitask neural network model after the preset number of iterations;
the identification module is used for acquiring personnel characteristics through the trained personnel multitask neural network model and identifying multidimensional characteristics;
and the construction module is used for collecting the multi-dimensional features and constructing the person integral portrait.
9. The person feature identification device according to claim 8, further comprising:
a sensor providing the person characteristic;
the control module acquires the personnel integral image and performs early warning;
and the studying and judging module is used for receiving the early warning information sent by the control module.
10. The person feature identification device according to claim 8, wherein the person multitasking neural network model is a parameter hard-shared HRNet-like neural network structure.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111435354.5A CN114332914A (en) | 2021-11-29 | 2021-11-29 | Personnel feature identification method, device and computer-readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111435354.5A CN114332914A (en) | 2021-11-29 | 2021-11-29 | Personnel feature identification method, device and computer-readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114332914A true CN114332914A (en) | 2022-04-12 |
Family
ID=81047663
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111435354.5A Pending CN114332914A (en) | 2021-11-29 | 2021-11-29 | Personnel feature identification method, device and computer-readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114332914A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529402A (en) * | 2016-09-27 | 2017-03-22 | 中国科学院自动化研究所 | Multi-task learning convolutional neural network-based face attribute analysis method |
| CN106815566A (en) * | 2016-12-29 | 2017-06-09 | 天津中科智能识别产业技术研究院有限公司 | A kind of face retrieval method based on multitask convolutional neural networks |
| CN108197701A (en) * | 2018-02-05 | 2018-06-22 | 哈工大机器人（合肥）国际创新研究院 | A kind of multi-task learning method based on RNN |
| CN108399649A (en) * | 2018-03-05 | 2018-08-14 | 中科视拓（北京）科技有限公司 | A kind of single picture three-dimensional facial reconstruction method based on cascade Recurrent networks |
| CN111862067A (en) * | 2020-07-28 | 2020-10-30 | 中山佳维电子有限公司 | Welding defect detection method and device, electronic equipment and storage medium |
| CN112530019A (en) * | 2020-12-11 | 2021-03-19 | 中国科学院深圳先进技术研究院 | Three-dimensional human body reconstruction method and device, computer equipment and storage medium |
-
2021
- 2021-11-29 CN CN202111435354.5A patent/CN114332914A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529402A (en) * | 2016-09-27 | 2017-03-22 | 中国科学院自动化研究所 | Multi-task learning convolutional neural network-based face attribute analysis method |
| CN106815566A (en) * | 2016-12-29 | 2017-06-09 | 天津中科智能识别产业技术研究院有限公司 | A kind of face retrieval method based on multitask convolutional neural networks |
| CN108197701A (en) * | 2018-02-05 | 2018-06-22 | 哈工大机器人（合肥）国际创新研究院 | A kind of multi-task learning method based on RNN |
| CN108399649A (en) * | 2018-03-05 | 2018-08-14 | 中科视拓（北京）科技有限公司 | A kind of single picture three-dimensional facial reconstruction method based on cascade Recurrent networks |
| CN111862067A (en) * | 2020-07-28 | 2020-10-30 | 中山佳维电子有限公司 | Welding defect detection method and device, electronic equipment and storage medium |
| CN112530019A (en) * | 2020-12-11 | 2021-03-19 | 中国科学院深圳先进技术研究院 | Three-dimensional human body reconstruction method and device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Singh et al. | A deeply coupled ConvNet for human activity recognition using dynamic and RGB images | |
| CN107169454B (en) | Face image age estimation method and device and terminal equipment thereof | |
| CN107506740B (en) | Human body behavior identification method based on three-dimensional convolutional neural network and transfer learning model | |
| AU2018100321A4 (en) | Person ReID method based on metric learning with hard mining | |
| CN106709461B (en) | Activity recognition method and device based on video | |
| Edel et al. | Binarized-blstm-rnn based human activity recognition | |
| Goodfellow et al. | Multi-digit number recognition from street view imagery using deep convolutional neural networks | |
| CN106570522B (en) | Object recognition model establishing method and object recognition method | |
| CN106909938B (en) | Visual angle independence behavior identification method based on deep learning network | |
| CN109376720A (en) | Classification of motion method based on artis space-time simple cycle network and attention mechanism | |
| CN109598220A (en) | A kind of demographic method based on the polynary multiple dimensioned convolution of input | |
| CN113111767A (en) | Fall detection method based on deep learning 3D posture assessment | |
| Rueda et al. | Learning attribute representation for human activity recognition | |
| CN105243154A (en) | Remote sensing image retrieval method and system based on significant point characteristics and spare self-encodings | |
| CN110929558A (en) | Pedestrian re-identification method based on deep learning | |
| CN112883931A (en) | Real-time true and false motion judgment method based on long and short term memory network | |
| CN112508269A (en) | Legal decision prediction method and system | |
| Defriani et al. | Recognition of regional traditional house in Indonesia using Convolutional Neural Network (CNN) method | |
| Chen | Image recognition technology based on neural network | |
| Li et al. | Multimodal gesture recognition using densely connected convolution and blstm | |
| CN110826459B (en) | Migratable campus violent behavior video identification method based on attitude estimation | |
| Hashim et al. | Deep transfer learning based human activity recognition by transforming IMU data to image domain using novel activity image creation method | |
| Al-Wesabi et al. | Design of optimal deep learning based human activity recognition on sensor enabled internet of things environment | |
| CN114332914A (en) | Personnel feature identification method, device and computer-readable storage medium | |
| CN113065468A (en) | Gait authentication method based on user coordinate system and GRU network |
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 |