CN111414812A - Human body attribute identification method, system, computer device and storage medium - Google Patents
Human body attribute identification method, system, computer device and storage medium Download PDFInfo
- Publication number
- CN111414812A CN111414812A CN202010138380.0A CN202010138380A CN111414812A CN 111414812 A CN111414812 A CN 111414812A CN 202010138380 A CN202010138380 A CN 202010138380A CN 111414812 A CN111414812 A CN 111414812A
- Authority
- CN
- China
- Prior art keywords
- human body
- attribute
- neural network
- image
- training
- 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 37
- 238000012549 training Methods 0.000 claims abstract description 169
- 238000013528 artificial neural network Methods 0.000 claims abstract description 151
- 238000001514 detection method Methods 0.000 claims abstract description 121
- 238000011176 pooling Methods 0.000 claims description 39
- 238000004590 computer program Methods 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 9
- 210000005036 nerve Anatomy 0.000 claims description 8
- 238000003062 neural network model Methods 0.000 abstract description 10
- 239000011159 matrix material Substances 0.000 description 39
- 230000008569 process Effects 0.000 description 11
- 230000001537 neural effect Effects 0.000 description 8
- 230000004913 activation Effects 0.000 description 6
- 238000013527 convolutional neural network Methods 0.000 description 6
- 238000009795 derivation Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000013507 mapping Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 239000003086 colorant Substances 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000009432 framing Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000002087 whitening effect Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008094 contradictory effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
Abstract
The embodiment of the invention provides a human body attribute identification method, which comprises the following steps: detecting the training image according to the human body detection model and the attribute identification model to obtain a human body image coordinate vector and a human body attribute vector; training the initial attribute recognition neural network according to the training image and the human body image coordinate vector to obtain a first attribute recognition neural network and a plurality of feature matrices; training the first attribute recognition neural network according to the plurality of feature matrices and the human body attribute vector to obtain a second attribute recognition neural network; performing combined training on the second attribute recognition neural network according to the training image, the human body image coordinate vector and the human body attribute vector to obtain a target attribute recognition neural network; and identifying the image to be identified by the neural network according to the target attribute so as to obtain the human body attribute value of the image to be identified. According to the embodiment of the invention, the human body detection task and the attribute identification task are combined into one neural network model, and the human body detection and the attribute identification are simultaneously realized by one neural network model, so that the efficiency of human body attribute identification is improved.
Description
Technical Field
The embodiment of the invention relates to the technical field of computers, in particular to a human body attribute identification method, a human body attribute identification system, computer equipment and a storage medium.
Background
The human body attribute identification is identification of gender, age, type and color of clothes and the like of people, and is widely applied to the fields of dangerous behavior early warning, traffic violation monitoring, industrial security and Automatic vending machines, Automatic Teller Machines (ATMs), target people locking in public places such as markets, stations and the like.
In the existing human body attribute identification scheme, a plurality of models are required to finish tasks of human body image extraction and attribute identification respectively, however, in the method for identifying the human body attributes by using the plurality of models, time loss of different degrees exists in joint operation among the models in the process of identifying the human body attributes. In addition, attribute identification is required for each detected human body, and the time loss in the human body attribute identification process cannot be kept stable.
Therefore, the scheme aims to solve the problems that the existing human body identification method cannot use a single model to identify attributes, and has time loss and low efficiency.
Disclosure of Invention
In view of this, embodiments of the present invention provide a human body attribute identification method, a human body attribute identification system, a computer device, and a computer readable storage medium, which can combine a human body detection task and an attribute identification task into a neural network model, and implement human body detection and attribute identification simultaneously by using one neural network model, thereby improving the efficiency of attribute identification.
The embodiment of the invention solves the technical problems through the following technical scheme:
a human body attribute identification method comprises the following steps:
acquiring a training image, and performing human body detection and attribute recognition on the training image according to a preset human body detection model and a preset attribute recognition model to obtain a human body image coordinate vector and a human body attribute vector of the training image;
establishing an initial attribute recognition neural network, inputting the training image and the human body image coordinate vector into the initial attribute recognition neural network, and performing human body detection training on the initial attribute recognition neural network to obtain a first attribute recognition neural network and a plurality of feature matrices, wherein the initial attribute recognition neural network comprises human body detection parameters and attribute recognition parameters;
inputting the plurality of feature matrixes and the human body attribute vector into the first attribute recognition neural network to train attribute recognition on the first attribute recognition neural network, so as to obtain a second attribute recognition neural network;
inputting the training image, the human body image coordinate vector and the human body attribute vector into the second attribute recognition neural network to perform combined training of human body detection and attribute recognition on the second attribute recognition neural network so as to obtain a target attribute recognition neural network;
and acquiring an image to be recognized, and recognizing the image to be recognized according to the target attribute recognition neural network so as to obtain a human body attribute value of the image to be recognized.
Further, the recognizing the training image according to a preset human body detection model and a preset attribute recognition model to obtain a human body image coordinate vector and a human body attribute vector of the training image includes:
human body detection is carried out on the training image according to the human body detection model so as to obtain a human body image and a human body image coordinate vector in the training image;
and performing attribute identification on the human body image according to the attribute identification model to obtain a human body attribute vector of the human body image.
Further, the performing attribute identification on the human body image according to the attribute identification model to obtain a human body attribute vector of the human body image includes:
inputting the human body image into the attribute identification model, so that each nerve layer in the attribute identification model performs convolution, pooling and full connection operations on the human body image, and a plurality of attribute identification results are obtained;
searching preset attributes and an attribute comparison table to obtain a target attribute value corresponding to each attribute identification result;
and combining the plurality of target attribute values into a human attribute vector of the human image.
Further, the inputting the training image and the human body image coordinate vector to the initial attribute recognition neural network for human body detection training of the initial attribute recognition neural network includes:
inputting the training image into the initial attribute recognition neural network for preprocessing, convolution and pooling to obtain the plurality of feature matrices and target human body image output, and updating the human body detection parameters according to the target human body image output and the human body image coordinate vector to obtain the first attribute recognition neural network.
Further, the inputting the plurality of feature matrices and the human body attribute vector into the first attribute recognition neural network to train attribute recognition on the first attribute recognition neural network, so as to obtain a second attribute recognition neural network includes:
calculating the plurality of feature matrices according to the attribute identification parameters to obtain target attribute values and output the target attribute values;
and updating the attribute identification parameters according to the target attribute value output and the human body attribute vector to obtain the second attribute identification neural network.
In order to achieve the above object, an embodiment of the present invention further provides a human body attribute identification system, including:
the human body detection and attribute identification module is used for acquiring a training image, and performing human body detection and attribute identification on the training image according to a preset human body detection model and a preset attribute identification model to obtain a human body image coordinate vector and a human body attribute vector of the training image;
the first training module is used for establishing an initial attribute recognition neural network, inputting the training image and the human body image coordinate vector into the initial attribute recognition neural network to carry out human body detection training on the initial attribute recognition neural network so as to obtain a first attribute recognition neural network and a plurality of feature matrices, wherein the initial attribute recognition neural network comprises human body detection parameters and attribute recognition parameters;
the second training module is used for inputting the plurality of feature matrixes and the human body attribute vector into the first attribute recognition neural network so as to train attribute recognition on the first attribute recognition neural network, and therefore a second attribute recognition neural network is obtained;
the joint training module is used for inputting the training image, the human body image coordinate vector and the human body attribute vector into the second attribute recognition neural network so as to carry out joint training of human body detection and attribute recognition on the second attribute recognition neural network, thereby obtaining a target attribute recognition neural network;
and the identification module is used for acquiring an image to be identified and identifying the image to be identified according to the target attribute identification neural network so as to obtain a human body attribute value of the image to be identified.
In order to achieve the above object, an embodiment of the present invention further provides a computer device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the steps of the human body attribute identification method as described above when executing the computer program.
In order to achieve the above object, an embodiment of the present invention further provides a computer-readable storage medium, in which a computer program is stored, the computer program being executable by at least one processor to cause the at least one processor to execute the steps of the human body attribute identification method as described above.
The human body attribute identification method, the human body attribute identification system, the computer equipment and the computer readable storage medium provided by the embodiment of the invention can combine the human body detection task and the attribute identification task into one neural network model, and realize human body detection and attribute identification simultaneously by using one neural network model, thereby improving the efficiency of attribute identification.
The invention is described in detail below with reference to the drawings and specific examples, but the invention is not limited thereto.
Drawings
FIG. 1 is a flowchart illustrating steps of a human body attribute identification method according to a first embodiment of the present invention;
FIG. 2 is a schematic diagram of a process module of a human body attribute identification system according to a second embodiment of the present invention;
fig. 3 is a schematic hardware structure diagram of a computer device according to a third embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Technical solutions between various embodiments may be combined with each other, but must be realized by those skilled in the art, and when the technical solutions are contradictory or cannot be realized, such a combination should not be considered to exist, and is not within the protection scope of the present invention.
Example one
Referring to fig. 1, a flowchart illustrating steps of a human body attribute identification method according to a first embodiment of the invention is shown. It is to be understood that the flow charts in the embodiments of the present method are not intended to limit the order in which the steps are performed. The following description is given by taking a computer device as an execution subject, specifically as follows:
step S100, a training image is obtained, and human body detection and attribute recognition are carried out on the training image according to a preset human body detection model and a preset attribute recognition model, so that a human body image coordinate vector and a human body attribute vector of the training image are obtained.
Specifically, before the initial attribute recognition neural network is established, the computer device first acquires a training image, and then detects the training image through a preset human body detection model and an attribute recognition model, so as to acquire a human body image in the training image and human body attribute values in the human body image, where the human body attribute values may be values of human body attributes such as gender, jacket style, jacket color, and clothing style, for example: (0,1,2,3), wherein the attribute label corresponding to 0 is male, the attribute label corresponding to 1 is long sleeve, the attribute label corresponding to 2 is white, and the attribute label corresponding to 3 is jeans. The coordinates of the human body image in the training image form the human body image coordinate vector, the human body attribute values in the human body image form the human body attribute vector, and the human body image coordinate vector and the human body attribute vector are used for training a newly-built initial attribute recognition neural network.
The training image may be a single frame image obtained by performing a frame decoding process on video data, where the video data may be obtained through a camera, a smart phone, a tablet computer, a personal computer, or other terminal devices. The preset human body detection model is an existing mature human body detection model, for example, based on ssd (single Shot multiple boxdetector) target detection framework or refledet target detection framework; the preset attribute recognition model is a trained attribute recognition model, and an attribute value corresponding to the preset attribute in the training image can be detected.
Illustratively, a single-frame image is obtained by performing de-framing processing on a monitoring video of a monitoring camera, and then the single-frame image is input into a preset human body detection model to obtain a human body image in the single-frame image and coordinates of the human body image; and then inputting the human body image into a preset attribute recognition model to obtain a human body attribute value.
In a preferred embodiment, a human body image region in a single frame image can be extracted through a refindedt target detection framework to obtain a human body image, and a coordinate vector (x, y, w, h) of the human body image relative to a preset coordinate system of the single frame image is obtained, wherein x represents an x coordinate of the upper left corner of the human body image in the preset coordinate system, y represents a y coordinate of the upper left corner of the human body image in the preset coordinate system, w represents a width of the human body image, and h represents a length of the human body image, the coordinates are all in units of pixels.
In another preferred embodiment, each human body attribute and the attribute value of each human body attribute can be predefined. For example, body attributes such as gender, age, jacket style, shirt style, jacket color, and shirt color, and attribute values of the body attributes may be defined as shown in table 1. The numbers in the table indicate the attribute values corresponding to the attribute tags, for example, the attribute value corresponding to the tag "male" is 0, and the attribute value corresponding to the tag "female" is 1. The database of the computer device is stored with a comparison table of each human body attribute and the corresponding attribute value, and the human body image is input into a preset attribute identification model to carry out attribute identification on the human body image. In specific implementation, the attribute identification model is a convolutional neural network model, and a plurality of attribute identification results are obtained after the human body image is subjected to operations such as convolution of a plurality of convolutional layers of the attribute identification model, pooling of pooling layers, full connection and the like. And then searching the attribute values in the comparison table to obtain a target attribute value corresponding to each attribute identification result, and finally combining a plurality of target attribute values into a human body attribute vector.
TABLE 1
Step S102, establishing an initial attribute recognition neural network, inputting the training image and the human body image coordinate vector into the initial attribute recognition neural network, and performing human body detection training on the initial attribute recognition neural network to obtain a first attribute recognition neural network and a plurality of feature matrices, wherein the initial attribute recognition neural network comprises human body detection parameters and attribute recognition parameters.
Specifically, an initial attribute recognition neural network is established on the basis of a convolutional neural network, and then training for human body detection is performed on the initial attribute recognition neural network by using the training image and the human body image coordinate vector. The initial attribute recognition neural network at least comprises an input layer, a convolutional layer, a pooling layer, an activation layer, a full-link layer and an output layer, and is divided into a human body detection branch and an attribute recognition branch, wherein the human body detection branch is used for extracting a human body image in a training image, and the attribute recognition branch is used for detecting human body attributes in the training image. The human detection branch further comprises human detection parameters, which refer to parameters of certain neural layers in the attribute recognition neural network that affect the human detection effect, such as: the width and height of the pooling window of the pooling layer, and/or the weight of the fully connected layer. The attribute identification branch further comprises attribute identification parameters, which refer to parameters of certain neural layers in the attribute identification neural network that affect the attribute identification effect, such as: the stride of the convolutional layer and the convolutional kernel. The human detection parameters and the attribute identification parameters are adjustable in the training process.
In this embodiment, the human detection branch is first trained, that is, the initial attribute recognition neural network is trained to detect the capability of the human image from the training image. When human body detection branch training is carried out, a plurality of feature matrixes are obtained after the training image is subjected to preprocessing of an initial attribute recognition neural network input layer, convolution of a convolution layer and pooling of a pooling layer, wherein the feature matrixes comprise a shallow feature matrix and a deep feature matrix, and the deep feature is obtained by convolution and pooling of the shallow feature matrix. When training of human body detection branches is carried out, the shallow feature matrix is used for coding detailed features such as human body textures, and the deep feature matrix is used for coding human body contour features. And the position of the human body in the image can be predicted through the common coding of the shallow characteristic matrix and the deep characteristic matrix. In this embodiment, after the calculation of the full connection layer and the adjustment of the human body detected branch parameters, the plurality of feature matrices output a target human body image on the output layer, where the data of the target human body image is represented by a target human body image coordinate vector (x, y, w, h). And finally, updating parameters of each nerve layer of the initial attribute recognition neural network and the parameters of the human body detected branch according to the target human body image coordinate vector and the human body image coordinate vector to obtain a first attribute recognition neural network.
In a preferred embodiment, the initial attribute-identifying neural network established based on the convolutional neural network may further be composed of an input layer, a first convolutional layer, a first pooling layer, a first activation layer, a second convolutional layer, a second pooling layer, a second activation layer, a full-link layer, and an output layer. When the human body detection branch is trained, the training image and the human body image coordinate vector are input into the input layer, and the training image is adjusted through parameters of each nerve layer and the human body detection branch to obtain the first attribute recognition nerve network and the plurality of feature matrices.
In another preferred embodiment, when the training image and the human body image coordinate vector are input to the initial attribute recognition neural network to perform human body detection training on the initial attribute recognition neural network, the training image may be input to the initial attribute recognition neural network to perform preprocessing, convolution and pooling so as to obtain the plurality of feature matrices and a target human body image output, and the human body detection parameter is updated according to the target human body image output and the human body image coordinate vector so as to obtain the first attribute recognition neural network.
Specifically, when the human body detection branch training is performed, the training image is input to an input layer of the initial attribute recognition neural network, and the input layer performs preprocessing such as mean value removal, normalization, Principal component analysis (PCV), whitening and the like on the training image, and then performs convolution of the first convolution layer to obtain a first feature matrix; obtaining a second feature matrix after the first feature matrix is subjected to pooling of the first pooling layer; the second feature matrix is subjected to convolution of the second convolution layer to obtain a third feature matrix; the third feature matrix is subjected to pooling of the second pooling layer to obtain a fourth feature matrix; and after the fourth characteristic matrix is calculated by the full connection layer and the human body detection branch parameters are adjusted, outputting a target human body image on an output layer, wherein the data expression form of the target human body image is a target human body image coordinate vector (x, y, w, h). And then calculating the coordinate vector of the target human body image and the coordinate vector of the human body image to obtain a training loss value. And performing partial derivative derivation and back propagation according to the training loss value to update the convolution kernel of the first convolution layer, the mapping parameter of the first pooling layer, the convolution kernel of the second convolution layer, the mapping parameter of the second pooling layer and the human body detection branch parameter so as to obtain a first attribute identification neural network.
Step S104, inputting the plurality of feature matrixes and the human body attribute vector into the first attribute recognition neural network to train attribute recognition on the first attribute recognition neural network, so as to obtain a second attribute recognition neural network.
Specifically, the plurality of feature matrices and the human body detection branch parameters are fixed, and then training of attribute identification is performed according to the plurality of feature matrices, wherein the human body attribute identification branch is divided into an overall attribute identification branch and a local attribute identification branch. In the plurality of feature matrices, a shallow feature matrix can represent local human attributes, such as jacket style and jacket color; the deep-level feature matrix may characterize the overall body attributes, such as gender and age. When training of attribute identification is carried out, the deep feature matrix is utilized to train the integral attribute branch, and integral attributes such as gender and age labels are learned and detected; and training the local attribute branches by using the shallow features, and learning and detecting local attributes such as upper and lower clothing styles, colors and the like. And after the first attribute recognition neural network is subjected to iterative training, obtaining the second attribute recognition neural network.
In a preferred embodiment, when the plurality of feature matrices and the human body attribute vector are input to the first attribute recognition neural network to train the first attribute recognition neural network for attribute recognition, the plurality of feature matrices are calculated according to the attribute recognition parameters to obtain target attribute value outputs, and the attribute recognition parameters are updated according to the target attribute value outputs and the human body attribute vector to obtain the second attribute recognition neural network.
Specifically, the plurality of feature matrices are calculated according to human body attribute identification branch parameters, wherein the attribute identification branch parameters are the attribute identification parameters. In this embodiment, the deep matrix characteristic is calculated according to the parameter of the overall attribute identification branch, and the shallow characteristic matrix is calculated according to the parameter of the local attribute identification branch, so as to obtain the target attribute value output. The target attribute value output is a discrimination value of the first attribute recognition neural network to each human body attribute corresponding to the human body attribute vector in the training image, and is a target output vector corresponding to the human body attribute vector. And then calculating a training loss value according to the target attribute value output and the human body attribute vector, and carrying out derivation according to the training loss value, thereby updating parameters of the human body attribute recognition branch.
Illustratively, the human body attribute vector is (0,1,2, 0,1, 1), the corresponding human body attribute labels are (gender male, jacket style short sleeve, under-coat style trousers, jacket color red, under-coat color gray, hat), and after the plurality of feature matrices are calculated by the parameters of the human body attribute identification branch, the obtained target attribute values are (0.91, 0.42, 0.73, 0.20, 0.84, 0.55), that is, the probability of gender male being 91%, the probability of jacket style short sleeve being 42%, the probability of under-coat style long sleeve being 73%, the probability of red being 20%, the probability of under-coat color gray being 84%, and the probability of hat being 55% in the training image. It should be noted that the discrimination probability of each human body attribute in the human body attribute vector is 1, that is, 100%. And then calculating a training loss value according to the target attribute value output and the human body attribute vector, and performing derivation according to the training loss value, thereby updating the parameters of the human body attribute recognition branch.
Step S106, inputting the training image, the human body image coordinate vector and the human body attribute vector into the second attribute recognition neural network to perform combined training of human body detection and attribute recognition on the second attribute recognition neural network, so as to obtain a target attribute recognition neural network.
Specifically, in the training of the attribute recognition in step S104, the plurality of feature matrices are fixed, so that part of information useful for the attribute recognition but useless for human body is filtered out in advance, and therefore, the training image, the human body image coordinate vector, and the human body attribute vector need to be input to the second attribute recognition neural network, and the joint training of human body detection and attribute recognition is performed, so as to obtain the optimal parameters of each neural layer, human body detection branch, and attribute recognition branch.
And S108, acquiring an image to be recognized, and recognizing the image to be recognized according to the target attribute recognition neural network so as to obtain a human body attribute value of the image to be recognized.
Specifically, a target attribute recognition neural network is obtained through human body detection and attribute recognition combined training, and parameters of each neural layer, human body detection branches and attribute recognition branches are adjusted to be in an optimal state. After the image to be recognized is obtained, the image to be recognized is input into the target attribute recognition neural network, and the human body attribute value of the image to be recognized can be obtained.
In the embodiment of the invention, the human body detection task and the attribute identification task can be combined into one neural network model, and the human body detection and the attribute identification can be simultaneously realized by using one neural network model, so that the efficiency of attribute identification is improved.
Example two
Referring to fig. 2, a schematic diagram of program modules of a human body attribute identification system according to a second embodiment of the invention is shown. It is to be understood that the flow charts in the embodiments of the present method are not intended to limit the order in which the steps are performed. In this embodiment, the human body attribute identification system 20 may include or be divided into one or more program modules, which are stored in a storage medium and executed by one or more processors to implement the present invention and implement the above-described human body attribute identification method. The program module referred to in the embodiments of the present invention refers to a series of computer program instruction segments capable of performing specific functions, and is more suitable for describing the execution process of the human body attribute recognition system 20 in the storage medium than the program itself. The following description will specifically describe the functions of the program modules of the present embodiment:
a human body detection and attribute recognition module 200, configured to obtain a training image, and perform human body detection and attribute recognition on the training image according to a preset human body detection model and a preset attribute recognition model to obtain a human body image coordinate vector and a human body attribute vector of the training image.
Specifically, before the initial attribute recognition neural network is established, the human body detection and attribute recognition module 200 first obtains a training image, and then detects the training image through a preset human body detection model and an attribute recognition model, so as to obtain a human body image in the training image and human body attribute values in the human body image, where the human body attribute values may be values of human body attributes such as gender, jacket style, jacket color, and clothing style, for example: (0,1,2,3), wherein the attribute label corresponding to 0 is male, the attribute label corresponding to 1 is long sleeve, the attribute label corresponding to 2 is white, and the attribute label corresponding to 3 is jeans. The coordinates of the human body image in the training image form the human body image coordinate vector, the human body attribute values in the human body image form the human body attribute vector, and the human body image coordinate vector and the human body attribute vector are used for training a newly-built initial attribute recognition neural network.
The training image may be a single frame image obtained by performing a frame decoding process on video data, where the video data may be obtained through a camera, a smart phone, a tablet computer, a personal computer, or other terminal devices. The preset human body detection model is an existing mature human body detection model, for example, based on ssd (single Shot multiple boxdetector) target detection framework or refledet target detection framework; the preset attribute recognition model is a trained attribute recognition model, and an attribute value corresponding to the preset attribute in the training image can be detected.
Illustratively, a single-frame image is obtained by performing de-framing processing on a monitoring video of a monitoring camera, and then the single-frame image is input into a preset human body detection model to obtain a human body image in the single-frame image and coordinates of the human body image; and then inputting the human body image into a preset attribute recognition model to obtain a human body attribute value.
In a preferred embodiment, the human body detection and attribute recognition module 200 may extract a human body image region in a single frame image through a RefineDet target detection framework to obtain a human body image, and obtain a coordinate vector (x, y, w, h) of the human body image relative to a preset coordinate system of the single frame image, where x represents an x coordinate of an upper left corner of the human body image in the preset coordinate system, y represents a y coordinate of the upper left corner of the human body image in the preset coordinate system, w represents a width of the human body image, and h represents a length of the human body image, and the coordinates are all in units of pixels, after the human body image is obtained, the human body image is input to a preset attribute recognition model for detection, and a human body attribute vector (0,1,2, 0,1, 1) is obtained, wherein an attribute label corresponding to the human body attribute vector is (male, long sleeve, shorts, white, black, no hat), the human body image coordinate vector and the human body attribute vector form a human body image coordinate attribute vector (x, y, w, h, 0,1,2, 0,1, 1) if multiple frame images are detected, the human body attribute vector can be represented by a human body coordinate vector ①.
In another preferred embodiment, the human body detection and attribute identification module 200 may further define each human body attribute and an attribute value of each human body attribute in advance. For example, body attributes such as gender, age, jacket style, shirt style, jacket color, and shirt color, and attribute values of the body attributes may be defined as shown in table 1. Wherein, the numbers in the table represent the attribute values corresponding to the attribute labels. The numbers in the table indicate the attribute values corresponding to the attribute tags, for example, the attribute value corresponding to the tag "male" is 0, and the attribute value corresponding to the tag "female" is 1. The database of the computer device is stored with a comparison table of each human body attribute and the corresponding attribute value, and the human body image is input into a preset attribute identification model to carry out attribute identification on the human body image. In specific implementation, the attribute identification model is a convolutional neural network model, and a plurality of attribute identification results are obtained after the human body image is subjected to operations such as convolution of a plurality of convolutional layers of the attribute identification model, pooling of pooling layers, full connection and the like. And then searching the attribute values in the comparison table to obtain a target attribute value corresponding to each attribute identification result, and finally combining a plurality of target attribute values into a human body attribute vector.
The first training module 202 is configured to establish an initial attribute recognition neural network, and input the training image and the human body image coordinate vector to the initial attribute recognition neural network to perform human body detection training on the initial attribute recognition neural network, so as to obtain a first attribute recognition neural network and a plurality of feature matrices, where the initial attribute recognition neural network includes a human body detection parameter and an attribute recognition parameter.
Specifically, the first training module 202 first establishes an initial attribute recognition neural network based on a convolutional neural network, and then performs human body detection training on the initial attribute recognition neural network by using the training image and the human body image coordinate vector. The initial attribute recognition neural network at least comprises an input layer, a convolutional layer, a pooling layer, an activation layer, a full-link layer and an output layer, and is divided into a human body detection branch and an attribute recognition branch, wherein the human body detection branch is used for extracting a human body image in a training image, and the attribute recognition branch is used for detecting human body attributes in the training image. The human detection branch further comprises human detection parameters, which refer to parameters of certain neural layers in the attribute recognition neural network that influence the human detection effect, such as the width and height of the pooling windows of the pooling layers, and/or the weight of the fully connected layers. The attribute identification branch further comprises attribute identification parameters, which refer to parameters of certain neural layers in the attribute identification neural network, such as the stride of the convolutional layer and the convolutional kernel, which influence the attribute identification effect. The human detection parameters and the attribute identification parameters are adjustable in the training process.
In this embodiment, the human detection branch is first trained, that is, the initial attribute recognition neural network is trained to detect the capability of the human image from the training image. When human body detection branch training is carried out, a plurality of feature matrixes are obtained after the training image is subjected to preprocessing of an initial attribute recognition neural network input layer, convolution of a convolution layer and pooling of a pooling layer, wherein the feature matrixes comprise a shallow feature matrix and a deep feature matrix, and the deep feature is obtained by convolution and pooling of the shallow feature matrix. When training of human body detection branches is carried out, the shallow feature matrix is used for coding detailed features such as human body textures, and the deep feature matrix is used for coding human body contour features. And the position of the human body in the image can be predicted through the common coding of the shallow characteristic matrix and the deep characteristic matrix. In this embodiment, after the calculation of the full connection layer and the adjustment of the human body detected branch parameters, the plurality of feature matrices output a target human body image on the output layer, where the data representation of the target human body image is the target human body image coordinate vector (x, y, w, h). And finally, updating parameters of each nerve layer of the initial attribute recognition neural network and the parameters of the human body detected branch according to the target human body image coordinate vector and the human body image coordinate vector to obtain a first attribute recognition neural network.
In a preferred embodiment, the initial attribute-identifying neural network established based on the convolutional neural network may further be composed of an input layer, a first convolutional layer, a first pooling layer, a first activation layer, a second convolutional layer, a second pooling layer, a second activation layer, a full-link layer, and an output layer. When the human body detection branch is trained, the training image and the human body image coordinate vector are input into the input layer, and the training image is adjusted through parameters of each nerve layer and the human body detection branch to obtain the first attribute recognition nerve network and the plurality of feature matrices.
In another preferred embodiment, when the first training module 202 inputs the training image and the human body image coordinate vector to the initial attribute recognition neural network for training human body detection on the initial attribute recognition neural network, the training image may be input to the initial attribute recognition neural network for preprocessing, convolution and pooling to obtain the plurality of feature matrices and a target human body image output, and the human body detection parameter is updated according to the target human body image output and the human body image coordinate vector to obtain the first attribute recognition neural network.
Specifically, when the human body detection branch training is performed, the training image is input to an input layer of the initial attribute recognition neural network, and the input layer performs preprocessing such as averaging, normalization, Principal Component Analysis (PCV), whitening, and the like on the training image; and then obtaining a first characteristic matrix through convolution of the first convolution layer. Obtaining a second feature matrix after the first feature matrix is subjected to pooling of the first pooling layer; the second feature matrix is subjected to convolution of the second convolution layer to obtain a third feature matrix; the third feature matrix is subjected to pooling of the second pooling layer to obtain a fourth feature matrix; and after the fourth characteristic matrix is calculated by the full connection layer and the human body detection branch parameters are adjusted, outputting a target human body image on an output layer. And the data expression form of the target human body image is a target human body image coordinate vector (x, y, w, h). The first training module 202 then calculates the target human body image coordinate vector and the human body image coordinate vector to obtain a training loss value. And performing partial derivative derivation and back propagation according to the training loss value to update the convolution kernel of the first convolution layer, the mapping parameter of the first pooling layer, the convolution kernel of the second convolution layer, the mapping parameter of the second pooling layer and the human body detection branch parameter so as to obtain a first attribute identification neural network.
A second training module 204, configured to input the plurality of feature matrices and the human body attribute vector to the first attribute recognition neural network to perform attribute recognition training on the first attribute recognition neural network, so as to obtain a second attribute recognition neural network.
Specifically, the second training module 204 fixes the plurality of feature matrices and the human body detection branch parameters, and then performs attribute recognition training according to the plurality of feature matrices, wherein the human body attribute recognition branch is further divided into an overall attribute recognition branch and a local attribute recognition branch. In the plurality of feature matrices, a shallow feature matrix can represent local human attributes, such as jacket style and jacket color; the deep-level feature matrix may characterize the overall body attributes, such as gender and age. When training of attribute identification is carried out, the deep feature matrix is utilized to train the integral attribute branch, and integral attributes such as gender and age labels are learned and detected; and training the local attribute branches by using the shallow features, and learning and detecting local attributes such as upper and lower clothing styles, colors and the like. And after the first attribute recognition neural network is subjected to iterative training, obtaining the second attribute recognition neural network.
In a preferred embodiment, when the second training module 204 inputs the feature matrices and the human body attribute vector to the first attribute recognition neural network for training attribute recognition of the first attribute recognition neural network, the feature matrices may be calculated according to the attribute recognition parameters to obtain target attribute value outputs, and the attribute recognition parameters may be updated according to the target attribute value outputs and the human body attribute vector to obtain the second attribute recognition neural network.
Specifically, the second training module 204 calculates the plurality of feature matrices according to the human body attribute identification branch parameters, wherein deep matrix features are calculated according to the parameters of the overall attribute identification branch, and shallow feature matrices are calculated according to the parameters of the local attribute identification branch, so as to obtain a target attribute value for output. The target attribute value output is a discrimination value of the first attribute recognition neural network to each human body attribute corresponding to the human body attribute vector in the training image, and is a target output vector corresponding to the human body attribute vector. And then calculating a training loss value according to the target attribute value output and the human body attribute vector, and carrying out derivation according to the training loss value, thereby updating parameters of the human body attribute recognition branch.
Illustratively, the human body attribute vector is (0,1,2, 0,1, 1), the corresponding human body attribute labels are (gender male, jacket style short sleeve, under-coat style trousers, jacket color red, under-coat color gray, hat), and after the plurality of feature matrices are calculated by the parameters of the human body attribute identification branch, the obtained target attribute values are (0.91, 0.42, 0.73, 0.20, 0.84, 0.55), that is, the probability of gender male being 91%, the probability of jacket style short sleeve being 42%, the probability of under-coat style long sleeve being 73%, the probability of red being 20%, the probability of under-coat color gray being 84%, and the probability of hat being 55% in the training image. It should be noted that the discrimination probability of each human body attribute in the human body attribute vector is 1, that is, 100%. And then calculating a training loss value according to the target attribute value output and the human body attribute vector, and performing derivation according to the training loss value, thereby updating the parameters of the human body attribute recognition branch.
A joint training module 206, configured to input the training image, the human body image coordinate vector, and the human body attribute vector to the second attribute recognition neural network to perform joint training of human body detection and attribute recognition on the second attribute recognition neural network, so as to obtain a target attribute recognition neural network.
Specifically, when the second training module 204 performs the training of attribute identification, the feature matrices are fixed, so that part of information which is useful for attribute identification but is useless for human body is filtered out in advance, and therefore the joint training module 206 needs to input the training image, the human body image coordinate vector and the human body attribute vector to the second attribute identification neural network, perform joint training of human body detection and attribute identification, and obtain the optimal parameters of each neural layer, human body detection branch and attribute identification branch.
The identification module 208 is configured to acquire an image to be identified, and identify the image to be identified according to the target attribute identification neural network, so as to obtain a human body attribute value of the image to be identified.
Specifically, a target attribute recognition neural network is obtained through human body detection and attribute recognition combined training, and parameters of each neural layer, human body detection branches and attribute recognition branches are adjusted to be in an optimal state. And acquiring an image to be recognized, inputting the image to be recognized into the target attribute recognition neural network, and obtaining a human body attribute value of the image to be recognized.
In the embodiment of the invention, the human body detection task and the attribute identification task can be combined into one neural network model, and the human body detection and the attribute identification can be simultaneously realized by using one neural network model, so that the efficiency of attribute identification is improved.
EXAMPLE III
Fig. 3 is a schematic diagram of a hardware architecture of a computer device according to a third embodiment of the present invention. In the present embodiment, the computer device 2 is a device capable of automatically performing numerical calculation and/or information processing in accordance with a preset or stored instruction. The computer device 2 may be a rack server, a blade server, a tower server or a rack server (including an independent server or a server cluster composed of a plurality of servers), and the like. As shown in FIG. 3, the computer device 2 includes, but is not limited to, at least a memory 21, a processor 22, a network interface 23, and a body attribute identification system 20, which may be communicatively coupled to each other via a system bus. Wherein:
in this embodiment, the memory 21 includes at least one type of computer-readable storage medium including a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a Read Only Memory (ROM), an Electrically Erasable Programmable Read Only Memory (EEPROM), a Programmable Read Only Memory (PROM), a magnetic memory, a magnetic disk, an optical disk, and the like. In some embodiments, the storage 21 may be an internal storage unit of the computer device 2, such as a hard disk or a memory of the computer device 2. In other embodiments, the memory 21 may also be an external storage device of the computer device 2, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), or the like provided on the computer device 2. Of course, the memory 21 may also comprise both internal and external memory units of the computer device 2. In this embodiment, the memory 21 is generally used for storing an operating system installed in the computer device 2 and various application software, such as the program codes of the human body attribute identification system 20 in the second embodiment. Further, the memory 21 may also be used to temporarily store various types of data that have been output or are to be output.
The network interface 23 may comprise a wireless network interface or a wired network interface, and the network interface 23 is generally used for establishing communication connection between the computer device 2 and other electronic apparatuses. For example, the network interface 23 is used to connect the computer device 2 to an external terminal through a network, establish a data transmission channel and a communication connection between the computer device 2 and the external terminal, and the like. The network may be a wireless or wired network such as an Intranet (Intranet), the Internet (Internet), a Global System of Mobile communication (GSM), Wideband Code Division Multiple Access (WCDMA), a 4G network, a 5G network, Bluetooth (Bluetooth), Wi-Fi, and the like.
It is noted that fig. 3 only shows the computer device 2 with components 20-23, but it is to be understood that not all shown components are required to be implemented, and that more or less components may be implemented instead.
In this embodiment, the human body attribute identification system 20 stored in the memory 21 may be further divided into one or more program modules, and the one or more program modules are stored in the memory 21 and executed by one or more processors (in this embodiment, the processor 22) to complete the present invention.
For example, fig. 2 shows a schematic diagram of program modules of the human body attribute recognition system 20, in this embodiment, the human body attribute recognition system 20 may be divided into a human body detection and attribute recognition module 200, a first training module 202, a second training module 204, a joint training module 206, and a recognition module 208. The program module referred to in the present invention refers to a series of computer program instruction segments capable of performing specific functions, and is more suitable than a program for describing the execution process of the human body attribute identification system 20 in the computer device 2. The specific functions of the program modules 200 and 208 have been described in detail in the second embodiment, and are not described herein again.
Example four
The present embodiment also provides a computer-readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, a server, an App application mall, etc., on which a computer program is stored, which when executed by a processor implements corresponding functions. The computer-readable storage medium of this embodiment is used for storing the human body attribute identification system 20, and when being executed by a processor, the computer-readable storage medium implements the human body attribute identification method of the first embodiment.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (10)
1. A human body attribute identification method is characterized by comprising the following steps:
acquiring a training image, and performing human body detection and attribute recognition on the training image according to a preset human body detection model and a preset attribute recognition model to obtain a human body image coordinate vector and a human body attribute vector of the training image;
establishing an initial attribute recognition neural network, inputting the training image and the human body image coordinate vector into the initial attribute recognition neural network, and performing human body detection training on the initial attribute recognition neural network to obtain a first attribute recognition neural network and a plurality of feature matrices, wherein the initial attribute recognition neural network comprises human body detection parameters and attribute recognition parameters;
inputting the plurality of feature matrixes and the human body attribute vector into the first attribute recognition neural network to train attribute recognition on the first attribute recognition neural network, so as to obtain a second attribute recognition neural network;
inputting the training image, the human body image coordinate vector and the human body attribute vector into the second attribute recognition neural network to perform combined training of human body detection and attribute recognition on the second attribute recognition neural network so as to obtain a target attribute recognition neural network;
and acquiring an image to be recognized, and recognizing the image to be recognized according to the target attribute recognition neural network so as to obtain a human body attribute value of the image to be recognized.
2. The human body attribute identification method according to claim 1, wherein the identifying the training image according to a preset human body detection model and a preset attribute identification model to obtain a human body image coordinate vector and a human body attribute vector of the training image comprises:
human body detection is carried out on the training image according to the human body detection model so as to obtain a human body image and a human body image coordinate vector in the training image;
and performing attribute identification on the human body image according to the attribute identification model to obtain a human body attribute vector of the human body image.
3. The human body attribute identification method according to claim 2, wherein the performing attribute identification on the human body image according to the attribute identification model to obtain a human body attribute vector of the human body image comprises:
inputting the human body image into the attribute identification model, so that each nerve layer in the attribute identification model performs convolution, pooling and full connection operations on the human body image, and a plurality of attribute identification results are obtained;
searching preset attributes and an attribute comparison table to obtain a target attribute value corresponding to each attribute identification result;
and combining the plurality of target attribute values into a human attribute vector of the human image.
4. The human body attribute recognition method according to claim 1, wherein the inputting the training image and the human body image coordinate vector to the initial attribute recognition neural network for human body detection training of the initial attribute recognition neural network comprises:
inputting the training image into the initial attribute recognition neural network for preprocessing, convolution and pooling to obtain the plurality of feature matrices and target human body image output, and updating the human body detection parameters according to the target human body image output and the human body image coordinate vector to obtain the first attribute recognition neural network.
5. The human body attribute identification method according to claim 1, wherein the inputting the plurality of feature matrices and the human body attribute vector to the first attribute recognition neural network for training attribute recognition of the first attribute recognition neural network to obtain a second attribute recognition neural network comprises:
calculating the plurality of feature matrices according to the attribute identification parameters to obtain target attribute values and output the target attribute values;
and updating the attribute identification parameters according to the target attribute value output and the human body attribute vector to obtain the second attribute identification neural network.
6. A human attribute identification system, comprising:
the human body detection and attribute identification module is used for acquiring a training image, and performing human body detection and attribute identification on the training image according to a preset human body detection model and a preset attribute identification model to obtain a human body image coordinate vector and a human body attribute vector of the training image;
the first training module is used for establishing an initial attribute recognition neural network, inputting the training image and the human body image coordinate vector into the initial attribute recognition neural network to carry out human body detection training on the initial attribute recognition neural network so as to obtain a first attribute recognition neural network and a plurality of feature matrices, wherein the initial attribute recognition neural network comprises human body detection parameters and attribute recognition parameters;
the second training module is used for inputting the plurality of feature matrixes and the human body attribute vector into the first attribute recognition neural network so as to train attribute recognition on the first attribute recognition neural network, and therefore a second attribute recognition neural network is obtained;
the joint training module is used for inputting the training image, the human body image coordinate vector and the human body attribute vector into the second attribute recognition neural network so as to carry out joint training of human body detection and attribute recognition on the second attribute recognition neural network, thereby obtaining a target attribute recognition neural network;
and the identification module is used for acquiring an image to be identified and identifying the image to be identified according to the target attribute identification neural network so as to obtain a human body attribute value of the image to be identified.
7. The human body attribute identification system of claim 6, wherein the human body detection and attribute identification module is further configured to:
human body detection is carried out on the training image according to the human body detection model so as to obtain a human body image and a human body image coordinate vector in the training image;
and performing attribute identification on the human body image according to the attribute identification model to obtain a human body attribute vector of the human body image.
8. The human attribute recognition system of claim 6, wherein the first training module is further configured to:
inputting the training image into the initial attribute recognition neural network for preprocessing, convolution and pooling to obtain the plurality of feature matrices and target human body image output, and updating the human body detection parameters according to the target human body image output and the human body image coordinate vector to obtain the first attribute recognition neural network.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the human body property recognition method according to any one of claims 1 to 5 when executing the computer program.
10. A computer-readable storage medium, in which a computer program is stored which is executable by at least one processor for causing the at least one processor to carry out the steps of the human body property recognition method according to any one of claims 1 to 5.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010138380.0A CN111414812A (en) | 2020-03-03 | 2020-03-03 | Human body attribute identification method, system, computer device and storage medium |
| PCT/CN2020/135080 WO2021174941A1 (en) | 2020-03-03 | 2020-12-10 | Physical attribute recognition method, system, computer device, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010138380.0A CN111414812A (en) | 2020-03-03 | 2020-03-03 | Human body attribute identification method, system, computer device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111414812A true CN111414812A (en) | 2020-07-14 |
Family
ID=71492794
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010138380.0A Pending CN111414812A (en) | 2020-03-03 | 2020-03-03 | Human body attribute identification method, system, computer device and storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN111414812A (en) |
| WO (1) | WO2021174941A1 (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111860347A (en) * | 2020-07-21 | 2020-10-30 | 海尔优家智能科技（北京）有限公司 | Article information adjusting method and device and electronic device |
| CN112182412A (en) * | 2020-11-26 | 2021-01-05 | 南京吉拉福网络科技有限公司 | Method, computing device, and computer storage medium for recommending physical examination items |
| CN112926427A (en) * | 2021-02-18 | 2021-06-08 | 浙江智慧视频安防创新中心有限公司 | Target user dressing attribute identification method and device |
| WO2021174941A1 (en) * | 2020-03-03 | 2021-09-10 | 平安科技（深圳）有限公司 | Physical attribute recognition method, system, computer device, and storage medium |
| WO2023030036A1 (en) * | 2021-08-30 | 2023-03-09 | 中科尚易健康科技(北京)有限公司 | Multi-camera-based human body meridian identification method and apparatus, and human body meridian conditioning device |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117173476A (en) * | 2023-09-05 | 2023-12-05 | 北京交通大学 | Single-source domain generalized pedestrian re-identification method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106599869A (en) * | 2016-12-22 | 2017-04-26 | 安徽大学 | Vehicle attribute identification method based on multi-task convolutional neural network |
| CN108875501A (en) * | 2017-11-06 | 2018-11-23 | 北京旷视科技有限公司 | Human body attribute recognition approach, device, system and storage medium |
| CN108875489A (en) * | 2017-09-30 | 2018-11-23 | 北京旷视科技有限公司 | Method for detecting human face, device, system, storage medium and capture machine |
| CN109271974A (en) * | 2018-11-16 | 2019-01-25 | 中山大学 | A kind of lightweight face joint-detection and recognition methods and its system |
| CN109522790A (en) * | 2018-10-08 | 2019-03-26 | 百度在线网络技术（北京）有限公司 | Human body attribute recognition approach, device, storage medium and electronic equipment |
| WO2019228089A1 (en) * | 2018-05-30 | 2019-12-05 | 腾讯科技（深圳）有限公司 | Human body attribute recognition method, apparatus, and device, and medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9514390B2 (en) * | 2014-12-17 | 2016-12-06 | Facebook, Inc. | Systems and methods for identifying users in media content based on poselets and neural networks |
| CN110059577B (en) * | 2019-03-26 | 2022-02-18 | 北京迈格威科技有限公司 | Pedestrian attribute information extraction method and device |
| CN110472611A (en) * | 2019-08-21 | 2019-11-19 | 图谱未来(南京)人工智能研究院有限公司 | Method, apparatus, electronic equipment and the readable storage medium storing program for executing of character attribute identification |
| CN111414812A (en) * | 2020-03-03 | 2020-07-14 | 平安科技（深圳）有限公司 | Human body attribute identification method, system, computer device and storage medium |
-
2020
- 2020-03-03 CN CN202010138380.0A patent/CN111414812A/en active Pending
- 2020-12-10 WO PCT/CN2020/135080 patent/WO2021174941A1/en active Application Filing
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106599869A (en) * | 2016-12-22 | 2017-04-26 | 安徽大学 | Vehicle attribute identification method based on multi-task convolutional neural network |
| CN108875489A (en) * | 2017-09-30 | 2018-11-23 | 北京旷视科技有限公司 | Method for detecting human face, device, system, storage medium and capture machine |
| CN108875501A (en) * | 2017-11-06 | 2018-11-23 | 北京旷视科技有限公司 | Human body attribute recognition approach, device, system and storage medium |
| WO2019228089A1 (en) * | 2018-05-30 | 2019-12-05 | 腾讯科技（深圳）有限公司 | Human body attribute recognition method, apparatus, and device, and medium |
| CN109522790A (en) * | 2018-10-08 | 2019-03-26 | 百度在线网络技术（北京）有限公司 | Human body attribute recognition approach, device, storage medium and electronic equipment |
| CN109271974A (en) * | 2018-11-16 | 2019-01-25 | 中山大学 | A kind of lightweight face joint-detection and recognition methods and its system |
Non-Patent Citations (3)
| Title |
|---|
| LIYING CHI ET AL: "End-to-End Spatial Transform Face Detection and Recognition", 《ARXIV:1703.10818V1》 * |
| LUWEI YANG ET AL: "Attribute recognition from adaptive Parts", 《 PROCEEDINGS OF THE BRITISH MACHINE VISION》 * |
| ZHU, JIANQING ET AL: "Multi-label convolutional neural network based pedestrian attribute classification", 《IMAGE AND VISION COMPUTING》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021174941A1 (en) * | 2020-03-03 | 2021-09-10 | 平安科技（深圳）有限公司 | Physical attribute recognition method, system, computer device, and storage medium |
| CN111860347A (en) * | 2020-07-21 | 2020-10-30 | 海尔优家智能科技（北京）有限公司 | Article information adjusting method and device and electronic device |
| CN112182412A (en) * | 2020-11-26 | 2021-01-05 | 南京吉拉福网络科技有限公司 | Method, computing device, and computer storage medium for recommending physical examination items |
| CN112926427A (en) * | 2021-02-18 | 2021-06-08 | 浙江智慧视频安防创新中心有限公司 | Target user dressing attribute identification method and device |
| WO2023030036A1 (en) * | 2021-08-30 | 2023-03-09 | 中科尚易健康科技(北京)有限公司 | Multi-camera-based human body meridian identification method and apparatus, and human body meridian conditioning device |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2021174941A1 (en) | 2021-09-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111414812A (en) | Human body attribute identification method, system, computer device and storage medium | |
| CN111428581B (en) | Face shielding detection method and system | |
| CN110147717B (en) | Human body action recognition method and device | |
| CN110197146B (en) | Face image analysis method based on deep learning, electronic device and storage medium | |
| CN105893920B (en) | Face living body detection method and device | |
| CN109685013B (en) | Method and device for detecting head key points in human body posture recognition | |
| WO2019033569A1 (en) | Eyeball movement analysis method, device and storage medium | |
| CN110866443B (en) | Portrait storage method, face recognition equipment and storage medium | |
| KR101835333B1 (en) | Method for providing face recognition service in order to find out aging point | |
| CN111914775B (en) | Living body detection method, living body detection device, electronic equipment and storage medium | |
| CN111462381A (en) | Access control method based on face temperature identification, electronic device and storage medium | |
| CN111626163B (en) | Human face living body detection method and device and computer equipment | |
| WO2021051547A1 (en) | Violent behavior detection method and system | |
| WO2019033570A1 (en) | Lip movement analysis method, apparatus and storage medium | |
| CN111666905B (en) | Model training method, pedestrian attribute identification method and related device | |
| US20220156959A1 (en) | Image processing device, image processing method, and recording medium in which program is stored | |
| WO2019033567A1 (en) | Method for capturing eyeball movement, device and storage medium | |
| CN111275040A (en) | Positioning method and device, electronic equipment and computer readable storage medium | |
| CN109871751A (en) | Attitude appraisal procedure, device and storage medium based on facial expression recognition | |
| WO2019061659A1 (en) | Method and device for removing eyeglasses from facial image, and storage medium | |
| CN111507324A (en) | Card frame identification method, device, equipment and computer storage medium | |
| CN113449704A (en) | Face recognition model training method and device, electronic equipment and storage medium | |
| CN114861241A (en) | Anti-peeping screen method based on intelligent detection and related equipment thereof | |
| WO2023185646A1 (en) | Systems and methods for image processing | |
| CN113989914B (en) | Security monitoring method and system based on face recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40032378 Country of ref document: HK |
|
| 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: 20200714 |