WO2020134858A1 - 人脸属性识别方法及装置、电子设备和存储介质 - Google Patents

人脸属性识别方法及装置、电子设备和存储介质 Download PDF

Info

Publication number
WO2020134858A1
WO2020134858A1 PCT/CN2019/121973 CN2019121973W WO2020134858A1 WO 2020134858 A1 WO2020134858 A1 WO 2020134858A1 CN 2019121973 W CN2019121973 W CN 2019121973W WO 2020134858 A1 WO2020134858 A1 WO 2020134858A1
Authority
WO
WIPO (PCT)
Prior art keywords
face
image
attribute
attributes
features
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.)
Ceased
Application number
PCT/CN2019/121973
Other languages
English (en)
French (fr)
Inventor
王飞
陈科宇
钱晨
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Sensetime Technology Development Co Ltd
Original Assignee
Beijing Sensetime Technology Development Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Sensetime Technology Development Co Ltd filed Critical Beijing Sensetime Technology Development Co Ltd
Priority to JP2021533534A priority Critical patent/JP2022515719A/ja
Priority to KR1020217020828A priority patent/KR20210113607A/ko
Publication of WO2020134858A1 publication Critical patent/WO2020134858A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/168—Feature extraction; Face representation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/09—Supervised learning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/172—Classification, e.g. identification
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/174—Facial expression recognition
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/178—Human faces, e.g. facial parts, sketches or expressions estimating age from face image; using age information for improving recognition

Definitions

  • the present disclosure relates to the field of image recognition technology, and in particular, to a face attribute recognition method and device, an electronic device, and a storage medium.
  • the present disclosure proposes a technical solution for face attribute recognition.
  • a face attribute recognition method including:
  • the attribute recognition result of the face in the image is determined according to the probabilistic prediction results of various face attributes.
  • a face attribute recognition device comprising:
  • Feature extraction module used to extract the features of the face in the image
  • a probabilistic prediction result obtaining module which is used to recognize preset multi-type face attributes according to the characteristics of the face, and obtain probabilistic prediction results of various face attributes;
  • the attribute recognition result obtaining module is used to determine the attribute recognition result of the face in the image according to the probabilistic prediction results of various face attributes.
  • an electronic device including:
  • Memory for storing processor executable instructions
  • the processor is configured to: execute the method described in any one of the above.
  • a computer-readable storage medium having computer program instructions stored thereon, which when executed by a processor implements the method described in any one of the above.
  • a computer program including computer readable code, and when the computer readable code runs in an electronic device, a processor in the electronic device executes a human face for performing the above Attribute recognition method.
  • the features of the face in the image can be extracted, and the preset multiple types of face attributes can be identified according to the features of the face to obtain the probabilistic prediction results of various types of face attributes.
  • the probability prediction result determines the attribute recognition result of the face in the image.
  • the embodiments of the present disclosure can simultaneously recognize multiple types of face attributes, and improve the recognition efficiency of face attribute recognition.
  • FIG. 1 shows a flowchart of a face attribute recognition method according to an embodiment of the present disclosure
  • FIG. 2 shows a flowchart of a face attribute recognition method according to an embodiment of the present disclosure
  • FIG. 3 shows a schematic diagram of a neural network training process in a face attribute recognition method according to an embodiment of the present disclosure
  • FIG. 4 shows a block diagram of a face attribute recognition device according to an embodiment of the present disclosure
  • Fig. 5 is a block diagram of an electronic device according to an exemplary embodiment
  • Fig. 6 is a block diagram of an electronic device according to an exemplary embodiment.
  • FIG. 1 shows a flowchart of a face attribute recognition method according to an embodiment of the present disclosure.
  • the face attribute recognition method includes:
  • step S10 the features of the face in the image are extracted.
  • the image may include a single image or a frame image in a video stream.
  • the image may include various types of images such as RGB (Red, Green, Blue, Red Green Blue) images.
  • the original high-dimensional feature data of the human face can be described with low-dimensional data, and the extracted low-dimensional features can effectively distinguish different human faces.
  • the features of the human face may include features of the global human face, and the feature information is extracted from the entire face image, and may reflect the overall face information.
  • the features of the human face may also include features of the human face based on locality, and feature information is extracted from partial pixels or partial regions of the facial image, which can reflect local details of the human face.
  • the present disclosure does not limit the types and specific contents of features of human faces. According to the needs of face attribute recognition, the features required by various face attributes can be extracted from the image.
  • step S20 the preset multiple types of face attributes are identified according to the characteristics of the face, and the probability prediction results of each face attribute are obtained.
  • the face attribute may include at least one of the following attributes: age, gender, ornaments, skin color, expression, motion, and ethnicity.
  • the decoration includes at least one of ordinary glasses, sunglasses, hats, and masks;
  • the expression includes at least one of happiness, sadness, crying, anger, and excitement;
  • the action includes: smoking, drinking water/drink , At least one of eating, makeup, and calling.
  • One or more attributes that make up a face attribute can be determined according to requirements.
  • face attributes can include: age and gender (such as middle-aged men or elderly men, etc.); or include age, gender, and ornaments (such as middle-aged men wearing hats, elderly women wearing glasses, etc.); or including age , Gender, skin color, facial expressions and movements (for example, middle-aged white males are laughing and calling, etc.).
  • multiple types of face attributes can be preset according to requirements.
  • the present disclosure does not limit the number and types of attributes among various types of face attributes.
  • three types of face attributes can be preset, of which face attribute 1 is: age 1, sex 1 and ornament 1; face attribute 2 is: age 2, sex 2, ornament 1, expression 1 and action 1 ; Face attribute 3 is: gender 3, ornaments 2, expression 2 and action 2.
  • the features of the face extracted in the image can be distinguished according to various types of preset face attributes to obtain features for various types of face attributes. After classifying the features of various face attributes, the probability prediction results of various face attributes are obtained.
  • Step S30 Determine the attribute recognition result of the face in the image according to the probability prediction result of each face attribute.
  • the attribute recognition result of the face includes: at least two attribute recognition results of the face.
  • the attribute recognition result of the face in the image may include the attribute recognition result of one or more types of faces.
  • a probability threshold may be preset, and one or more types of face attributes whose probability prediction result is greater than the probability threshold may be determined as the attribute recognition result of the face in the image.
  • the features of the face in the image can be extracted, and the preset multiple types of face attributes can be identified according to the features of the face to obtain the probabilistic prediction results of various types of face attributes.
  • the probability prediction result determines the attribute recognition result of the face in the image.
  • the embodiments of the present disclosure can simultaneously recognize multiple types of face attributes, and improve the recognition efficiency of face attribute recognition.
  • the image includes an image of a road where the vehicle is located and/or an image inside the vehicle taken by an on-board camera installed on the vehicle, and the face in the image includes pedestrians on the road where the vehicle is located and other vehicles
  • the attribute recognition result includes an attribute recognition result that meets at least one of the following conditions: setting age, setting gender, setting decoration, setting Set skin color, set expressions, set actions, set race.
  • a vehicle-mounted camera may be set on the vehicle, a vehicle-mounted camera may be set to simultaneously capture the image in the vehicle and the road where the vehicle is located, or two cameras may be set separately, one to capture the image in the vehicle and one to capture the vehicle
  • This road image is not limited in this disclosure.
  • the car camera can be a visible light camera, an infrared camera or a binocular camera, etc., and a car camera can be selected according to needs and application scenarios.
  • the on-vehicle camera may capture an image of the road on which the vehicle is located.
  • the image of the road on which the vehicle is located may include the face of a person walking on the road, or the face of the driver or passenger of another vehicle on the road.
  • the method in the embodiments of the present disclosure may be used to determine the attribute recognition results of the faces of pedestrians or other vehicle drivers on the road.
  • the correspondence between the attribute recognition result of the face and the preset first target object can be preset, and the face with dangerous actions (phone calls, makeup, etc.) and unstable emotions can be determined as the first target Object.
  • the first target object may be an object with risk factors, which may cause interference to driving or a dangerous situation.
  • pedestrian A is attribute recognition result 1 (child, running)
  • pedestrian B is attribute recognition result 2 (elderly, wearing sunglasses, standing)
  • driver A in vehicle A is attribute recognition result 3 (calling)
  • passenger in vehicle B Attribute recognition result 4 happy, smoking.
  • the pedestrian in A and the driver in vehicle A may be determined as the first target object according to the correspondence between the attribute recognition result and the preset first target object.
  • the in-vehicle camera may capture an in-vehicle image
  • the in-vehicle image may include the vehicle driver and passengers.
  • the method in the embodiments of the present disclosure may be used to determine the attribute recognition result of the face of the driver and/or passenger of the vehicle.
  • the attribute recognition result that meets at least one of the following conditions can be determined according to requirements: setting the age, setting the gender, setting the ornament, setting the skin color, setting the expression, and setting the human face as the first target object.
  • the set age may include children and the elderly
  • the set gender may include women or men
  • the set ornaments may include sunglasses
  • the set expressions may include anger and sadness
  • the set actions may include smoking and makeup.
  • the above setting conditions may be combined to obtain an attribute recognition result that satisfies the setting conditions, and the face that meets the setting conditions is determined as the first target object.
  • the correspondence between the attribute recognition result of the face and the second target object can be preset, and the face with dangerous actions (calling, makeup, sleeping, etc.) and having unstable emotions can be determined as the second target.
  • vehicle driver is attribute recognition result 1 (eating)
  • passenger A is attribute recognition result 2 (wearing sunglasses, happy)
  • passenger B is attribute recognition result 3 (calling)
  • passenger C is attribute recognition result 4 (happy ,smokes).
  • the driver of the vehicle may be determined as the second target object according to the correspondence between the attribute recognition result and the preset second target object. Correspondence between different attribute recognition results and the second target object may be set for the driver and passenger of the vehicle.
  • the method further includes:
  • the face attribute recognition of the image in the vehicle and the image of the road where the vehicle is located may be used for intelligent driving, assisted driving, a safe driving system, and the like. It can be used to send prompt information to the vehicle driver or vehicle monitoring personnel, and can also be used to send the set driving control information to the vehicle driving control system, so that the vehicle can perform the corresponding driving action.
  • the prompt information may be sent according to the determined first target object.
  • the disclosure does not limit the specific implementation manner and content of the prompt information.
  • voice prompts such as "Please pay attention to children running ahead" and "Please don't call the driver, please be safe” can be sent.
  • the set driving control information may be sent to the vehicle driving control system to control the vehicle to perform at least one driving action of deceleration, steering or braking. It is also possible to send prompt information and set driving control information. This disclosure does not limit this.
  • the in-vehicle camera and/or image of the road where the vehicle is located is captured by the on-board camera installed on the vehicle to obtain the faces of pedestrians, other vehicle occupants and/or vehicle occupants on the road where the vehicle is located Attribute recognition result.
  • the attribute recognition result provided by this implementation can be used to improve the safety of intelligent driving and assisted driving.
  • the image includes a surveillance image captured by a surveillance camera installed at a surveillance location, and the surveillance location includes any of the following locations: shopping malls, schools, large venues, roads, and office buildings;
  • the attribute recognition result includes the attribute recognition result that meets at least one of the following conditions: setting age, setting gender, setting decoration, setting skin color, setting expression, setting action, setting race.
  • a monitoring camera may be set at a set position of a monitoring location such as a shopping mall, a school, a large venue, a road surface, and an office building. For example, it can be at entrances, exits, key roads, and key locations.
  • the surveillance camera can capture surveillance images inside or outside the mall, inside the school playground or classroom, and inside large venues.
  • the monitoring image may include multiple faces.
  • the method in the embodiment of the present disclosure may be used to determine the attribute recognition result of the face in the monitoring image. Wherein, the attribute recognition result satisfying the set condition is determined to be the second target object, and the relevant description determined to be the first target object may be referred to, which will not be repeated here.
  • a surveillance camera can be installed in a shopping mall to take a surveillance image to obtain the attribute recognition result of the face of a worker or buyer in the shopping mall.
  • You can set up surveillance cameras in the classroom or playground of the school to take surveillance images to get the attribute recognition results of the faces of teachers, students or personnel outside the school.
  • the attribute recognition result of the face in the monitoring image can be used for security work, or to directionally send setting information.
  • the method further includes:
  • the correspondence between the attribute recognition result of the face and the preset second target object may be preset, and a face with unstable emotion may be determined as the second target object, It is also possible to determine a face such as a female who meets the recognition result of the set attribute as the second target object.
  • a face such as a female who meets the recognition result of the set attribute as the second target object.
  • the student can be used as the second target object, and the broadcast message of "please leave classmates out of the playground" is sent in the broadcast.
  • the attribute recognition result of the face of the mall monitoring image female buyers can be used as the second target object, and setting information such as cosmetic ads can be played on the large screen of the mall, and male buyers can be used as the second target object in the mall.
  • the setting information of car advertising light is played on the screen. It is also possible to send prompt information to the monitoring personnel at the monitoring location according to the second target object, to prompt the second target object to be a dangerous object or a monitoring object.
  • the present disclosure does not limit the implementation manner and information content of setting information and prompt information.
  • the surveillance image set at the surveillance location captures the surveillance image to obtain the attribute recognition result of the face in the surveillance image.
  • the attribute recognition result provided by this implementation can be used to improve the security of the monitoring location or send setting information to the monitoring location.
  • FIG. 2 shows a flowchart of a face attribute recognition method according to an embodiment of the present disclosure.
  • step S10 in the face attribute recognition method includes:
  • Step S11 Perform convolution processing on the image to extract convolution features of the human face.
  • the convolution kernel is usually a two-dimensional matrix of m*n, and the convolution kernel and the step size can be used to convolve the image blocks at different positions in the image and extract it from the image.
  • features for example, features such as the outline and color depth of the target object in the image can be extracted as the convolution features of the human face.
  • Step S20 includes:
  • Step S21 Extract features of various face attributes in the convolution features of the face to obtain attribute features of each face attribute.
  • the convolution features of the extracted face include features of multiple types of face attributes.
  • attribute features for various face attributes can be obtained.
  • the convolution features of the face extracted in image 1 include (feature information 1, feature information 2, ... feature information 100), and the convolution features of the extracted face can be based on the preset three types of face attributes Make a distinction to get the attribute features for face attribute 1 (feature information 1, feature information 2...feature information 50), the attribute features for face attribute 2 (feature information 20, feature information 21...feature information 80), Attribute features for face attribute 3 (feature information 1, feature information 2, ... feature information 100).
  • Step S22 Perform classification processing on the attribute characteristics of each of the face attributes to confirm the probability prediction result of each face attribute.
  • classification processing can be performed according to attribute characteristics of various face attributes to obtain probabilistic prediction results of various face attributes. For example, the probability prediction result of image 1 for face attribute 1 is 10%, the probability prediction result for face attribute 2 is 80%, and the probability prediction result for face attribute 3 is 70%. The probability threshold is 50%, then the face attribute recognition results in image 1 are: face attribute 2 and face attribute 3.
  • the convolution features in the image can be extracted, and after obtaining the attribute features for various face attributes according to the convolution features, the attribute features of various face attributes can be classified and processed to obtain various types of faces Probability prediction results of attributes. Convolutional features and classification processing can make the probability prediction results of face attributes more accurate.
  • Step S11 includes:
  • the convolutional layer of the neural network performs convolution processing on the image to extract convolutional features of the image.
  • Step S21 includes:
  • the features of each face attribute in the convolution features of the face are extracted through the fully connected layer of the neural network to obtain the attribute features of each face attribute.
  • Step S22 includes:
  • the classification feature of each face attribute is classified through the classification layer of the neural network to confirm the probability prediction result of each face attribute.
  • the neural network may include an input layer, an intermediate layer, and an output layer.
  • the input layer, the middle layer and the output layer can include several neurons, and directional connections with variable weights can be used between each neuron.
  • the neural network achieves the purpose of establishing a relationship model between analog input and output through repeated learning and training of known information, and by gradually adjusting and changing the connection weight of neurons.
  • the trained neural network can use the relationship model between the simulated input and output to detect the input information and give the output information corresponding to the input information.
  • the neural network may include multiple convolutional layers, fully connected layers, classification layers, and so on. You can convolve the image through the convolution layer of the neural network to extract the convolution features of the image, use the fully connected layer to extract the attribute features for various face attributes, and then convert the various face attributes. The attribute features are classified by the classification layer to obtain the probabilistic prediction results of various face attributes.
  • the image can be input into a neural network, and the convolutional layer, the fully connected layer, and the classification layer in the neural network can be used to extract the convolutional features in the image, and the attribute features for various face attributes can be obtained according to the convolutional features. , And then classify the attribute features to obtain the probabilistic prediction results of various face attributes in the image.
  • the powerful processing ability of the neural network can make the recognition efficiency of face attribute recognition higher and the recognition result more accurate.
  • the neural network performs pre-supervised training based on a sample image set, where the sample images in the sample image set include annotation information of multiple types of face attributes.
  • a sample image set composed of sample images including multiple types of face attributes may be used to supervise and train the neural network.
  • the sample image set includes N sample images, and each sample image may include annotation information of multiple types of face attributes.
  • the face in the sample image 1 includes the annotation information of three types of face attributes, which are the annotation information of the first type face attribute, the fourth type face attribute, and the eighth type face attribute, respectively.
  • the face includes the labeling information of two types of face attributes, which are the labeling information of the type 2 face attribute and the type 4 face attribute...
  • the neural network can be trained using a sample image set composed of sample images including multiple types of face attributes.
  • a sample image set composed of sample images including multiple types of face attributes may be used to supervise and train the neural network.
  • the training efficiency of the neural network can be improved, so that the trained neural network can recognize multiple types of face attributes at the same time.
  • the training process of the neural network includes:
  • the prediction results of each face attribute in the sample image are obtained through a neural network.
  • the second loss is determined according to the first loss of each face attribute.
  • the sample image may be input into the neural network to obtain prediction results of various face attributes in the sample image.
  • the sample image 1 includes three types of face attribute labeling information, namely face attribute 1 tag information, face attribute 2 tag information, face attribute 3 tag information, face attribute 4 tag information and face Labeling information for attribute 5.
  • the annotation information of each face attribute can be obtained according to the actual face attributes in the sample image. For example, when the face in the sample image does not have face attribute 3, the annotation information of face attribute 3 may be 0 or empty.
  • the preset face attributes in the neural network include 5, which are face attribute 1 to face attribute 5, respectively.
  • the prediction results of the sample image 1 for various face attributes are: the prediction result of the face attribute 1 is 80%, the prediction result of the face attribute 2 is 40%, the prediction result of the face attribute 3 is 60%, the person The prediction result of face attribute 4 is 10%, and the prediction result of face attribute 5 is 20%.
  • the first loss of various face attributes can be calculated according to the prediction results of various face attributes and the tag information of the face attributes of various face attributes.
  • the first loss of the target face attribute can be calculated according to the preset loss function, the prediction result of the target face attribute and the label information of the target face attribute.
  • the target face attribute is any face attribute.
  • the first loss of various face attributes in the sample image may be added to obtain the second loss.
  • step S300 includes: according to the first loss of the target face attribute and the weight of the target face attribute, determining the weight loss of the target face attribute, the target face
  • the attributes include any one of the face attributes; the second loss of the neural network is determined according to the weight loss of various face attributes.
  • different weights can be set for various face attributes.
  • five types of face attributes are preset, in which the weight value of face attribute 1 is K1, and the weight value of face attribute 2 is K2...
  • the weight loss of the target face attribute can be obtained by multiplying the first loss of each face attribute and its corresponding weight. After adding the weight loss of each face attribute, the second loss is obtained.
  • the gradient of the second loss can be propagated back to the neural network to adjust the network parameters of the neural network.
  • the sample images in the sample image set can be input to the neural network in turn, and the iterative training can be performed on the neural network until the preset convergence conditions or the preset training times are reached, and the training of the neural network is completed.
  • the sample image can be input into a neural network to obtain prediction results of various face attributes, and the first loss of the prediction result of each face attribute and its labeled information can be determined separately.
  • the second loss is determined, and the network parameters of the neural network are adjusted according to the second loss.
  • the parameters of the neural network are adjusted according to the loss determined by the loss of various face attributes. The correlation between various face attributes is fully reflected in the parameter adjustment process of the neural network, which can improve the accuracy of the neural network in identifying multiple types of face attributes.
  • FIG. 3 shows a schematic diagram of a neural network training process in a face attribute recognition method according to an embodiment of the present disclosure.
  • the neural network training process includes:
  • the convolutional layer of the neural network extracts the features of the face.
  • the fully connected layer of the neural network obtains attribute characteristics for various face attributes according to the preset face attributes. As shown in FIG. 3, the attribute features of face attribute 1 and the attribute features of face attribute 2... the attribute features of face attribute N are obtained.
  • the classification layer of the neural network classifies the attribute features of various face attributes, and obtains prediction results of various face attributes. According to the prediction results of various face attributes and the annotation information of the sample images, the first loss of various face attributes is obtained.
  • the second loss is obtained.
  • the training method of the neural network further includes:
  • it can be used to train the original image of the neural network, add noise such as horizontal flipping, random cropping, translation, and random perturbation in the color space to obtain a noise image of the original image, and combine the noise image and the original image All are used as sample images to obtain sample image sets.
  • noise such as horizontal flipping, random cropping, translation, and random perturbation in the color space
  • the number of sample images can be increased, and the generalization ability of the neural network can be improved.
  • step S10 includes: adjusting the resolution of the image to obtain a first processed image; and extracting features of a human face in the first processed image.
  • the image with the adjusted resolution can be input to the neural network to perform face attribute recognition.
  • the resolution of the image can be adjusted to 224 ⁇ 224, 128 ⁇ 128, 96 ⁇ 96, 64 ⁇ 64 according to requirements.
  • the lower the resolution of the image the faster the neural network can recognize face attributes.
  • the resolution of images can be reduced to improve the rate of face attribute recognition.
  • the rate of facial attribute recognition by the neural network can be increased by adjusting the resolution of the image.
  • the method further includes:
  • the convolutional layer of the neural network performs convolution processing on the image to extract convolutional features of the image, including:
  • the convolutional layer of the adjusted neural network performs convolution processing on the image to extract convolutional features of the image.
  • the number of channels of the characteristic of the neural network can be adjusted, including adjusting the number of channels of the convolution kernel of the convolutional layer in the neural network. It is possible to increase the recognition speed of facial attributes of neural networks by reducing the number of channels.
  • FIG. 4 shows a block diagram of a face attribute recognition apparatus according to an embodiment of the present disclosure.
  • the face attribute recognition apparatus includes:
  • the feature extraction module 100 is used to extract the features of the face in the image
  • the probability prediction result acquisition module 200 is used to recognize preset multiple types of face attributes according to the characteristics of the face, and obtain probability prediction results of various types of face attributes;
  • the attribute recognition result obtaining module 300 is used to determine the attribute recognition result of the face in the image according to the probabilistic prediction results of various face attributes.
  • the attribute recognition result of the face includes:
  • At least two attribute recognition results of the face At least two attribute recognition results of the face.
  • the face attribute includes at least one of the following attributes: age, gender, ornaments, skin color, expression, motion, and ethnicity.
  • the ornament includes at least one of ordinary glasses, sunglasses, hats, and masks; the expression includes: at least one of happiness, sadness, crying, anger, and excitement; the Actions include: at least one of smoking, drinking water/drinks, eating, makeup, and making phone calls.
  • the image includes an image of a road where the vehicle is located and/or an image inside the vehicle, which is captured by an on-board camera installed on the vehicle,
  • the faces in the image include pedestrians on the road where the vehicle is located, faces of other vehicle occupants and/or faces of occupants in the vehicle,
  • the attribute recognition result includes an attribute recognition result that satisfies at least one of the following conditions: setting age, setting gender, setting ornaments, setting skin color, setting expression, setting action, setting ethnicity.
  • the device further includes:
  • a first target object determination module configured to determine the first target object according to the attribute recognition result of the human face in the image
  • the driving control module is configured to send prompt information and/or control commands to the vehicle control system according to the determined first target object, so that the vehicle performs one of the following driving actions: speed maintenance, direction maintenance, deceleration, steering, brake.
  • the image includes a surveillance image captured by a surveillance camera installed at a surveillance location
  • the surveillance location includes any of the following locations: shopping malls, schools, large venues, road surfaces, and office buildings;
  • the attribute recognition result includes an attribute recognition result that satisfies at least one of the following conditions: setting age, setting gender, setting ornaments, setting skin color, setting expression, setting action, setting ethnicity.
  • the device further includes:
  • a first target object determination module configured to determine a second target object according to the attribute recognition result of the human face in the monitoring image
  • a monitoring control module configured to send setting information at the monitoring location according to the determined second target object, or
  • the feature extraction module is used to:
  • the probability prediction result acquisition module includes:
  • the face attribute feature extraction submodule is used to extract the features of various face attributes in the convolution features of the face and obtain the attribute features of various face attributes;
  • the probabilistic prediction result acquisition sub-module is used for classifying the attribute characteristics of the various face attributes to confirm the probabilistic prediction results of various face attributes.
  • the feature extraction module is used to:
  • the convolutional layer of the neural network performs convolution processing on the image to extract convolutional features of the image
  • the face attribute feature extraction sub-module is used to:
  • the probabilistic prediction result acquisition sub-module is used to:
  • the classification features of the various face attributes are classified through the classification layer of the neural network to confirm the probabilistic prediction results of various face attributes.
  • the neural network performs pre-supervised training based on a sample image set, where the sample images in the sample image set include annotation information of multiple types of face attributes.
  • the device further includes a training module, the training module is used for training the neural network, and the training module is used for:
  • the determining the second loss of the neural network according to the first loss of various face attributes includes:
  • the second loss of the neural network is determined according to the weight loss of the various face attributes.
  • the training module is further configured to: at least one of flip, crop, translate, and add noise to the first image to obtain a noise image of the first image;
  • the first image and the noise image are determined as sample images to obtain the sample image set.
  • the feature extraction module includes:
  • a first processed image acquisition module used to adjust the resolution of the image to obtain a first processed image
  • the first feature extraction sub-module is used to extract features of the human face in the first processed image.
  • the device further includes:
  • a channel adjustment module used to adjust the number of channels of the characteristic of the neural network to obtain an adjusted neural network
  • the feature extraction module is also used to:
  • the convolutional layer of the adjusted neural network performs convolution processing on the image to extract convolutional features of the image.
  • the functions provided by the apparatus provided by the embodiments of the present disclosure or the modules contained therein may be used to perform the methods described in the above method embodiments.
  • An embodiment of the present disclosure also proposes a computer-readable storage medium on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the above method is implemented.
  • the computer-readable storage medium may be a non-volatile computer-readable storage medium.
  • An embodiment of the present disclosure also provides an electronic device, including: a processor; a memory for storing processor-executable instructions; wherein, the processor is configured as the above method.
  • the electronic device may be provided as a terminal, server, or other form of device.
  • Fig. 5 is a block diagram of an electronic device 800 according to an exemplary embodiment.
  • the electronic device 800 may be a terminal such as a mobile phone, a computer, a digital broadcasting terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, and a personal digital assistant.
  • the electronic device 800 may include one or more of the following components: a processing component 802, a memory 804, a power supply component 806, a multimedia component 808, an audio component 810, an input/output (I/O) interface 812, and a sensor component 814 , ⁇ 816.
  • the processing component 802 generally controls the overall operations of the electronic device 800, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations.
  • the processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps in the above method.
  • the processing component 802 may include one or more modules to facilitate interaction between the processing component 802 and other components.
  • the processing component 802 may include a multimedia module to facilitate interaction between the multimedia component 808 and the processing component 802.
  • the memory 804 is configured to store various types of data to support operation at the electronic device 800. Examples of these data include instructions for any application or method for operating on the electronic device 800, contact data, phone book data, messages, pictures, videos, etc.
  • the memory 804 may be implemented by any type of volatile or nonvolatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable and removable Programmable read only memory (EPROM), programmable read only memory (PROM), read only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.
  • SRAM static random access memory
  • EEPROM electrically erasable programmable read only memory
  • EPROM erasable and removable Programmable read only memory
  • PROM programmable read only memory
  • ROM read only memory
  • magnetic memory flash memory
  • flash memory magnetic disk or optical disk.
  • the power supply component 806 provides power to various components of the electronic device 800.
  • the power component 806 may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the electronic device 800.
  • the multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user.
  • the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user.
  • the touch panel includes one or more touch sensors to sense touch, swipe, and gestures on the touch panel. The touch sensor may not only sense the boundary of the touch or sliding action, but also detect the duration and pressure related to the touch or sliding operation.
  • the multimedia component 808 includes a front camera and/or a rear camera. When the electronic device 800 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera may receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have focal length and optical zoom capabilities.
  • the audio component 810 is configured to output and/or input audio signals.
  • the audio component 810 includes a microphone (MIC).
  • the microphone is configured to receive an external audio signal.
  • the received audio signal may be further stored in the memory 804 or transmitted via the communication component 816.
  • the audio component 810 further includes a speaker for outputting audio signals.
  • the I/O interface 812 provides an interface between the processing component 802 and a peripheral interface module.
  • the peripheral interface module may be a keyboard, a click wheel, or a button. These buttons may include, but are not limited to: home button, volume button, start button, and lock button.
  • the sensor component 814 includes one or more sensors for providing the electronic device 800 with status assessment in various aspects.
  • the sensor component 814 can detect the on/off state of the electronic device 800, and the relative positioning of the components, for example, the component is the display and keypad of the electronic device 800, and the sensor component 814 can also detect the electronic device 800 or the electronic device 800.
  • the position of the component changes, the presence or absence of user contact with the electronic device 800, the orientation or acceleration/deceleration of the electronic device 800, and the temperature change of the electronic device 800.
  • the sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact.
  • the sensor component 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications.
  • the sensor component 814 may further include an acceleration sensor, a gyro sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
  • the communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices.
  • the electronic device 800 can access a wireless network based on a communication standard, such as WiFi, 2G, or 3G, or a combination thereof.
  • the communication component 816 receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel.
  • the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication.
  • the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.
  • RFID radio frequency identification
  • IrDA infrared data association
  • UWB ultra-wideband
  • Bluetooth Bluetooth
  • the electronic device 800 may be used by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field Programming gate array (FPGA), controller, microcontroller, microprocessor or other electronic components are used to implement the above method.
  • ASICs application specific integrated circuits
  • DSPs digital signal processors
  • DSPDs digital signal processing devices
  • PLDs programmable logic devices
  • FPGA field Programming gate array
  • controller microcontroller, microprocessor or other electronic components are used to implement the above method.
  • a non-volatile computer-readable storage medium is also provided, for example, a memory 804 including computer program instructions, which can be executed by the processor 820 of the electronic device 800 to complete the above method.
  • An embodiment of the present disclosure also provides a computer program product, including computer readable code.
  • a processor in the device executes instructions for implementing the method provided in any of the above embodiments.
  • the computer program product may be implemented in hardware, software, or a combination thereof.
  • the computer program product is embodied as a computer storage medium, and in another alternative embodiment, the computer program product is embodied as a software product, such as a software development kit (Software Development Kit, SDK), etc. Wait.
  • a software development kit Software Development Kit, SDK
  • Fig. 6 is a block diagram of an electronic device 1900 according to an exemplary embodiment.
  • the electronic device 1900 may be provided as a server.
  • the electronic device 1900 includes a processing component 1922, which further includes one or more processors, and memory resources represented by the memory 1932 for storing instructions executable by the processing component 1922, such as application programs.
  • the application programs stored in the memory 1932 may include one or more modules each corresponding to a set of instructions.
  • the processing component 1922 is configured to execute instructions to perform the above method.
  • the electronic device 1900 may also include a power component 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to the network, and an input output (I/O) interface 1958 .
  • the electronic device 1900 can operate an operating system based on the memory 1932, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM or the like.
  • a non-volatile computer-readable storage medium is also provided, for example, a memory 1932 including computer program instructions, which can be executed by the processing component 1922 of the electronic device 1900 to complete the above method.
  • the present disclosure may be a system, method, and/or computer program product.
  • the computer program product may include a computer-readable storage medium loaded with computer-readable program instructions for causing the processor to implement various aspects of the present disclosure.
  • the computer-readable storage medium may be a tangible device that can hold and store instructions used by the instruction execution device.
  • the computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing.
  • Computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), and erasable programmable read only memory (EPROM (Or flash memory), static random access memory (SRAM), portable compact disk read-only memory (CD-ROM), digital versatile disk (DVD), memory stick, floppy disk, mechanical encoding device, such as a computer on which instructions are stored
  • RAM random access memory
  • ROM read only memory
  • EPROM erasable programmable read only memory
  • SRAM static random access memory
  • CD-ROM compact disk read-only memory
  • DVD digital versatile disk
  • memory stick floppy disk
  • mechanical encoding device such as a computer on which instructions are stored
  • the convex structure in the hole card or the groove and any suitable combination of the above.
  • the computer-readable storage medium used here is not to be interpreted as a transient signal itself, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (for example, optical pulses through fiber optic cables), or through wires The transmitted electrical signal.
  • the computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to various computing/processing devices, or downloaded to an external computer or external storage device through a network, such as the Internet, a local area network, a wide area network, and/or a wireless network.
  • the network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and/or edge servers.
  • the network adapter card or network interface in each computing/processing device receives computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing/processing device .
  • the computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or in one or more programming languages Source code or object code written in any combination.
  • the programming languages include object-oriented programming languages such as Smalltalk, C++, etc., and conventional procedural programming languages such as "C" language or similar programming languages.
  • Computer readable program instructions can be executed entirely on the user's computer, partly on the user's computer, as an independent software package, partly on the user's computer and partly on a remote computer, or completely on the remote computer or server carried out.
  • the remote computer may be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (eg, using an Internet service provider to pass the Internet connection).
  • electronic circuits such as programmable logic circuits, field programmable gate arrays (FPGAs), or programmable logic arrays (PLA), are personalized by utilizing the state information of computer-readable program instructions.
  • Computer-readable program instructions are executed to implement various aspects of the present disclosure.
  • These computer-readable program instructions can be provided to the processor of a general-purpose computer, special-purpose computer, or other programmable data processing device, thereby producing a machine that causes these instructions to be executed by the processor of a computer or other programmable data processing device A device that implements the functions/actions specified in one or more blocks in the flowchart and/or block diagram is generated.
  • the computer-readable program instructions may also be stored in a computer-readable storage medium. These instructions cause the computer, programmable data processing apparatus, and/or other devices to work in a specific manner. Therefore, the computer-readable medium storing the instructions includes An article of manufacture that includes instructions to implement various aspects of the functions/acts specified in one or more blocks in the flowchart and/or block diagram.
  • the computer-readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other equipment, so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other equipment to produce a computer-implemented process , So that the instructions executed on the computer, other programmable data processing device, or other equipment implement the functions/acts specified in one or more blocks in the flowchart and/or block diagram.
  • each block in the flowchart or block diagram may represent a part of a module, program segment, or instruction that contains one or more Executable instructions.
  • the functions marked in the blocks may also occur in an order different from that marked in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and sometimes they can also be executed in reverse order, depending on the functions involved.
  • each block in the block diagrams and/or flowcharts, and combinations of blocks in the block diagrams and/or flowcharts can be implemented with a dedicated hardware-based system that performs specified functions or actions Or, it can be realized by a combination of dedicated hardware and computer instructions.

Landscapes

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

Abstract

一种人脸属性识别方法及装置、电子设备和存储介质。所述方法包括：提取图像中人脸的特征（S10）；根据所述人脸的特征进行预设的多类人脸属性的识别，得到各类人脸属性的概率预测结果（S20）；根据各类人脸属性的概率预测结果确定所述图像中人脸的属性识别结果（S30）。该方法可以同时进行多类人脸属性的识别，提高人脸属性识别的识别效率。

Description

人脸属性识别方法及装置、电子设备和存储介质
本公开要求在2018年12月29日提交中国专利局、申请号为201811645459.1、申请名称为“人脸属性识别方法及装置、电子设备和存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本公开中。
技术领域
本公开涉及图像识别技术领域，尤其涉及一种人脸属性识别方法及装置、电子设备和存储介质。
背景技术
在传统的人脸属性识别技术中，当需要对人脸进行多个属性的识别时，需要进行多个单属性人脸的识别，再将多个单属性人脸识别的结果进行整合。由于没有考虑各人脸属性之间的相关性，多属性人脸识别的结果准确度差，且计算速度低。
发明内容
本公开提出了一种人脸属性识别技术方案。
根据本公开的一方面，提供了一种人脸属性识别方法，包括：
提取图像中人脸的特征；
根据所述人脸的特征进行预设的多类人脸属性的识别，得到各类人脸属性的概率预测结果；
根据各类人脸属性的概率预测结果确定所述图像中人脸的属性识别结果。
根据本公开的一方面，提供了一种人脸属性识别装置，所述装置包括：
特征提取模块，用于提取图像中人脸的特征；
概率预测结果获取模块，用于根据所述人脸的特征进行预设的多类人脸属性的识别，得到各类人脸属性的概率预测结果；
属性识别结果获取模块，用于根据各类人脸属性的概率预测结果确定所述图像中人脸的属性识别结果。
根据本公开的一方面，提供了一种电子设备，包括：
处理器；
用于存储处理器可执行指令的存储器；
其中，所述处理器被配置为：执行上述任意一项所述的方法。
根据本公开的一方面，提供了一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述任意一项所述的方法。
根据本公开的一方面，提供了一种计算机程序，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行用于执行上述的人脸属性识别方法。
在本公开实施例中，可以提取图像中人脸的特征，根据人脸的特征进行预设的多类人脸属性的识别，得到各类人脸属性的概率预测结果，根据各类人脸属性的概率预测结果确定图像中人脸的属性识别结果。本公开实施例可以同时进行多类人脸属性的识别，提高人脸属性识别的识别效率。
应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，而非限制本公开。
根据下面参考附图对示例性实施例的详细说明，本公开的其它特征及方面将变得清楚。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，这些附图示出了符合本公开的实施例，并与说明书一起用于说明本公开的技术方案。
图1示出根据本公开实施例的人脸属性识别方法的流程图；
图2示出根据本公开实施例的人脸属性识别方法的流程图；
图3示出根据本公开实施例的人脸属性识别方法中神经网络的训练流程示意图；
图4示出根据本公开实施例的人脸属性识别装置的框图；
图5是根据一示例性实施例示出的一种电子设备的框图；
图6是根据一示例性实施例示出的一种电子设备的框图。
具体实施方式
以下将参考附图详细说明本公开的各种示例性实施例、特征和方面。附图中相同的附图标记表示功能相同或相似的元件。尽管在附图中示出了实施例的各种方面，但是除非特别指出，不必按比例绘制附图。
在这里专用的词“示例性”意为“用作例子、实施例或说明性”。这里作为“示例性”所说明的任何实施例不必解释为优于或好于其它实施例。
本文中术语“和/或”，仅仅是一种描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本文中术语“至少一种”表示多种中的任意一种或多种中的至少两种的任意组合，例如，包括A、B、C中的至少一种，可以表示包括从A、B和C构成的集合中选择的任意一个或多个元素。
另外，为了更好地说明本公开，在下文的具体实施方式中给出了众多的具体细节。本领域技术人员应当理解，没有某些具体细节，本公开同样可以实施。在一些实例中，对于本领域技术人员熟知的方法、手段、元件和电路未作详细描述，以便于凸显本公开的主旨。
图1示出根据本公开实施例的人脸属性识别方法的流程图，如图1所示，所述人脸属性识别方法，包括：
步骤S10，提取图像中人脸的特征。
在一种可能的实现方式中，图像可以包括单幅的图像，也可以包括视频流中的帧图像。图像可以包括RGB(Red，Green，Blue，红绿蓝)图像等各种类型的图像。
在一种可能的实现方式中，可以将人脸原始的高维特征数据用低维数据进行描述，提取出的低维特征可以有效的区分不同的人脸。人脸的特征可以包括基于全局的人脸的特征，其特征信息从人脸图像的整体上提取，可以反映整体的人脸信息。人脸的特征还可以包括基于局部的人脸的特征，特征信息从人脸图像的部分像素或部分区域提取，能够体现人脸的局部细节。本公开不限定人脸的特征的类型和具体的内容。可以根据人脸属性识别的需求，在图像中提取各类人脸属性所需的特征。
步骤S20，根据所述人脸的特征进行预设的多类人脸属性的识别，得到各人脸属性的概率预测结果。
在一种可能的实现方式中，所述人脸属性可以包括以下属性中的至少一种：年龄、性别、装饰物、肤色、表情、动作、种族。所述装饰物包括普通眼镜、墨镜、帽子、口罩中的至少一种；所述表情包括：高兴、悲伤、哭泣、愤怒、激动中的至少一种；所述动作包括：抽烟、喝水/饮料、进食、化妆、打电话中的至少一种。可以根据需求确定组成一种人脸属性的一种属性或多种属性。
例如，人脸属性可以包括：年龄和性别(例如中年男性或老年男性等)；或包括年龄、性别和装饰物(例如戴帽子的中年男性、戴眼镜的老年女性等)；或包括年龄、性别、肤色、表情和动作(例如中年白人男性正在大笑打电话等)。
在一种可能的实现方式中，可以根据需求预设多类人脸属性。本公开不限定各类人脸属性中属性的数量和类型。例如，可以预设三类人脸属性，其中，人脸属性1为：年龄1、性别1和装饰物1；人脸属性2为：年龄2、性别2、装饰物1、表情1和动作1；人脸属性3为：性别3、装饰物2、表情2和动作2。
在一种可能的实现方式中，可以将在图像中提取的人脸的特征，按照预设的各类人脸属性进行区分后，得到针对各类人脸属性的特征。再将针对各类人脸属性的特征进行分类处理后，得到各类人脸属性的概率预测结果。
步骤S30，根据各人脸属性的概率预测结果确定所述图像中人脸的属性识别结果。
在一种可能的实现方式中，所述人脸的属性识别结果，包括：所述人脸的至少两个属性识别结果。
在一种可能的实现方式中，图像中人脸的属性识别结果，可以包括一类或多类人脸的属性识别结果。可以预设概率阈值，将概率预测结果大于概率阈值的一类或多类人脸属性，确定为图像中人脸的属性识别结果。
在本实施例中，可以提取图像中人脸的特征，根据人脸的特征进行预设的多类人脸属性的识别，得到各类人脸属性的概率预测结果，根据各类人脸属性的概率预测结果确定图像中人脸的属性识别结果。本公开实施例可以同时进行多类人脸属性的识别，提高人脸属性识别的识别效率。
在一种可能的实现方式中，所述图像包括利用设置在车辆上的车载摄像头拍摄的车辆所在道路图像和/或车内图像，所述图像中的人脸包括车辆所在道路上行人、其他车辆驾乘人员的人脸和/或车内驾乘人员的人脸，所述属性识别结果，包括至少满足以下条件之一的属性识别结果：设定年龄、设定性别、设定装饰物、设定肤色、设定表情、设定动作、设定种族。
在一种可能的实现方式中，可以在车辆上设置车载摄像头，可以设置一个车载摄像头同时拍摄车内图像和车辆所在道路图像，也可以分别设置两个摄像头，一个拍摄车内图像，一个拍摄车辆所在道路图像，本公开对此不做限定。车载摄像头可以为可见光摄像头、红外摄像头或双目摄像头等，可以根据需求和应用场景选用车载摄像头。
在一种可能的实现方式中，车载摄像头可以拍摄得到车辆所在道路图像，车辆所在道路图像中可以包括道路上行人的人脸，也可以包括道路上其他车辆驾驶员或乘客的人脸。可以利用本公开实施例中的方法，确定道路上行人或其他车辆驾乘人员人脸的属性识别结果。可以预设人脸的属性识别结果和预设的第一目标对象之间的对应关系，可以将有危险性动作(打电话、化妆等)、具有不稳定情绪的人脸，确定为第一目标对象。第一目标对象可以是具有危险因素的对象，对驾驶会产生干扰或出现危险情况。
例如，行人A为属性识别结果1(小孩、跑步)、行人B为属性识别结果2(老人、戴墨镜、站立)、车辆A中驾驶员为属性识别结果3(打电话)、车辆B中乘客为属性识别结果4(高兴、抽烟)。可以根据属性识别结果和预设的第一目标对象之间的对应关系，确定行人A和车辆A中驾驶员为第一目标对象。
在一种可能的实现方式中，车载摄像头可以拍摄得到车内图像，车内图像中可以包括车辆驾驶员和乘客。可以利用本公开实施例中的方法，确定车辆驾驶员和/或乘客的人脸的属性识别结果。可以根据需求，确定至少满足以下条件之一的属性识别结果：设定年龄、设定性别、设定装饰物、设定肤色、设定表情、设定动作的人脸作为第一目标对象。其中，设定年龄可以包括小孩、老人，设定性别可以包括女性或男性，设定装饰物可以包括墨镜，设定表情可以包括愤怒、悲伤等，设定动作可以包括抽烟、化妆等。可以将上述各设定条件进行组合，得到满足设定条件的属性识别结果，并将满足设定条件的人脸确定为第一目标对象。例如，可以预设人脸的属性识别结果和第二目标对象之间的对应关系，可以将有危险性动作(打电话、化妆、睡觉等)、具有不稳定情绪的人脸，确定为第二目标对象。
例如，车辆驾驶员为属性识别结果1(吃东西)、乘客A为属性识别结果2(戴墨镜、高兴)、乘客B为属性识别结果3(打电话)、乘客C为属性识别结果4(高兴、抽烟)。可以根据属性识别结果和预设的第二目标对象之间的对应关系，确定车辆驾驶员为第二目标对象。可以为车辆驾驶员和乘客设置不同的属性识别结果和第二目标对象之间的对应关系。
在一种可能的实现方式中，所述方法还包括：
根据所述图像中人脸的属性识别结果确定第一目标对象；
根据所述确定出的第一目标对象发送提示信息和/或向车辆控制系统发送控制指令，以使车辆执行以下驾驶动作之一：速度保持、方向保持、减速、转向、刹车。
在一种可能的实现方式中，车内图像和车辆所在道路图像的人脸属性识别可以用于智能驾驶、辅助驾驶、安全驾驶系统等。可以用于向车辆驾驶员或车辆监控人员发送提示信息，也可以用于向车辆驾驶控制系统发送设定的驾驶控制信息，以使车辆执行相应的驾驶动作。
在一种可能的实现方式中，可以根据确定出的第一目标对象，发送提示信息。本公开不限定提示信息的具体实现方式及提示内容。例如，可以发送“前方有跑动的小孩请注意”、“驾驶员请不要打电话，请注意安全”的语音提示。也可以根据确定出的第一目标对象，向车辆驾驶控制系统发送设定的驾驶控制信息控制车辆进行减速、转向或刹车中的至少一项驾驶动作。也可以发送提示信息并发送设定的驾驶控制信息。本公开对此不做限定。
在本实施例中，通过在车辆上设置的车载摄像头拍摄车内图像和/或车辆所在道路图像，得到车辆所在道路上的行人、其他车辆驾乘人员和/或车内驾乘人员人脸的属性识别结果。本实施提供的属 性识别结果可以用于提高智能驾驶、辅助驾驶等的安全性。
在一种可能的实现方式中，所述图像包括利用设置在监控地点的监控摄像头拍摄的监控图像，所述监控地点包括以下任一地点：商场、学校、大型场馆、路面、办公楼宇；所述属性识别结果，包括至少满足以下条件之一的属性识别结果：设定年龄、设定性别、设定装饰物、设定肤色、设定表情、设定动作、设定种族。
在一种可能的实现方式中，可以在商场、学校、大型场馆、路面、办公楼宇等监控地点的设定位置设置监控摄像头。例如，可以在出入口、关键道路、重点地点。监控摄像头可以拍摄得到商场内部或外部、学校操场或教室内、大型场馆内部的监控图像。上述监控地点出入人员众多，监控图像中可以包括多个人脸。可以利用本公开实施例中的方法，确定监控图像中的人脸的属性识别结果。其中，满足设定条件的属性识别结果确定为第二目标对象，可以参照确定为第一目标对象的相关描述，不再赘述。
例如，可以在商场设置监控摄像头拍摄监控图像，得到商场内工作人员或购买者人脸的属性识别结果。可以在学校的教室内或操场设置监控摄像头拍摄监控图像，得到教师、学生或校外人员人脸的属性识别结果。监控图像中人脸的属性识别结果可以用于安全保卫工作、或定向发送设定信息。
在一种可能的实现方式中，所述方法还包括：
根据所述监控图像中人脸的属性识别结果确定第二目标对象；
根据所述确定出的第二目标对象在所述监控地点发送设定信息，或
根据所述确定出的第二目标对象向监控人员发送提示信息。
在一种可能的实现方式中，可以预设人脸的属性识别结果和预设的第二目标对象之间的对应关系，可以将有具有不稳定情绪的人脸，确定为第二目标对象，也可以将女性等满足设定属性识别结果的人脸确定为第二目标对象。例如，在放学时间根据学校监控图像人脸的属性识别结果，可以将学生作为第二目标对象，在广播中发送“请各位同学离开操场”的广播信息。在商场监控图像人脸的属性识别结果中，可以将女性购买者作为第二目标对象，在商场大屏幕上播放化妆品广告等设定信息，可以将男性购买者作为第二目标对象，在商场大屏幕上播放播放汽车广告灯设定信息。也可以根据第二目标对象，向监控地点的监控人员发送提示信息，提示第二目标对象为危险对象或监控对象等。本公开不限定设定信息活提示信息的实现方式和信息内容。
在本实施例中，通过在监控地点设置的监控摄像头拍摄监控图像，得到监控图像中人脸的属性识别结果。本实施提供的属性识别结果可以用于提高监控地点的安全性或向监控地点发送设定信息等。
图2示出根据本公开实施例的人脸属性识别方法的流程图，如图2所示，所述人脸属性识别方法中步骤S10，包括：
步骤S11，对所述图像进行卷积处理，以提取所述人脸的卷积特征。
在一种可能的实现方式中，卷积核通常为m*n的二维矩阵，可以利用预设的卷积核和步长对图像中不同位置的图像块进行卷积处理，在图像中提取特定的特征，例如可以提取图像中目标对象的轮廓、颜色深浅等特征，作为人脸的卷积特征。
步骤S20，包括：
步骤S21，提取所述人脸的卷积特征中各类人脸属性的特征，得到各人脸属性的属性特征。
在一种可能的实现方式中，在提取到的人脸的卷积特征中，包括多类人脸属性的特征。可以根据预设的人脸属性，在提取到的卷积特征中，得到针对各类人脸属性的属性特征。例如，在图像1中提取的人脸的卷积特征包括(特征信息1、特征信息2……特征信息100)，可以根据预设的三类人脸属性，将提取的人脸的卷积特征进行区分，得到针对人脸属性1的属性特征(特征信息1、特征信息2……特征信息50)、针对人脸属性2的属性特征(特征信息20、特征信息21……特征信息80)、针对人脸属性3的属性特征(特征信息1、特征信息2……特征信息100)。
步骤S22，对各所述人脸属性的属性特征进行分类处理，以确认各人脸属性的概率预测结果。
在一种可能的实现方式中，可以根据各类人脸属性的属性特征进行分类处理，得到各类人脸属性的概率预测结果。例如，图像1对于人脸属性1的概率预测结果为10％，针对人脸属性2的概率预测结 果为80％，针对人脸属性3的概率预测结果为70％。概率阈值为50％，则图像1中人脸的属性识别结果为：人脸属性2和人脸属性3。
在本实施例中，可以提取图像中的卷积特征，并根据卷积特征得到针对各类人脸属性的属性特征后，对各类人脸属性的属性特征进行分类处理，得到各类人脸属性的概率预测结果。卷积特征和分类处理，可以使得人脸属性的概率预测结果更加准确。
在一种可能的实现方式中，所述人脸属性识别方法中：
步骤S11，包括：
经神经网络的卷积层对所述图像进行卷积处理，以提取所述图像的卷积特征。
步骤S21，包括：
经所述神经网络的全连接层提取所述人脸的卷积特征中各人脸属性的特征，得到各人脸属性的属性特征。
步骤S22，包括：
经所述神经网络的分类层对各所述人脸属性的属性特征进行分类处理，以确认各人脸属性的概率预测结果。
在一种可能的实现方式中，神经网络可以包括输入层、中间层和输出层。输入层、中间层和输出层都可以包括若干神经元，各神经元之间可以用带可变权重的有向连接。神经网络通过对已知信息的反复学习训练，通过逐步调整改变神经元连接权重的方法，达到建立模拟输入输出之间关系模型的目的。训练好的神经网络可以利用模拟好的输入输出之间的关系模型，检测输入信息，并给出与输入信息对应的输出信息。
在一种可能的实现方式中，神经网络可以包括多个卷积层、全连接层和分类层等。可以经神经网络的卷积层对图像进行卷积处理，提取图像的卷积特征，将卷积特征利用全连接层提取针对各类人脸属性的属性特征后，再将各类人脸属性的属性特征经分类层进行分类处理，得到各类人脸属性的概率预测结果。
在本实施例中，可以将图像输入神经网络，利用神经网络中卷积层、全连接层和分类层，提取图像中的卷积特征，根据卷积特征得到针对各类人脸属性的属性特征，再将属性特征进行分类处理，得到图像中各类人脸属性的概率预测结果。神经网络强大的处理能力，可以使得人脸属性识别的识别效率更高、识别结果更加准确。
在一种可能的实现方式中，所述神经网络基于样本图像集预先监督训练完成，其中，所述样本图像集中的样本图像包括有多类人脸属性的标注信息。
在一种可能的实现方式中，可以利用包括多类人脸属性的样本图像组成的样本图像集，对神经网络进行监督训练。例如，样本图像集中包括N个样本图像，各样本图像可以包括多类人脸属性的标注信息。例如，样本图像1中的人脸包括三类人脸属性的标注信息，分别为第1类人脸属性、第4类人脸属性和第8类人脸属性的标注信息，样本图像2中的人脸包括两类人脸属性的标注信息，分别为第2类人脸属性和第4类人脸属性的标注信息……。可以利用包括多类人脸属性的样本图像组成的样本图像集，对神经网络进行训练。
在本实施例中，可以利用包括多类人脸属性的样本图像组成的样本图像集，对神经网络进行监督训练。可以提高神经网络的训练效率，使得训练完成的神经网络能够同时识别多类人脸属性。
在一种可能的实现方式中，所述神经网络的训练过程包括：
经神经网络分别获得所述样本图像中各人脸属性的预测结果。
分别确定各人脸属性的预测结果与其标注信息的第一损失。
根据各人脸属性的第一损失确定第二损失。
根据所述第二损失调整神经网络的网络参数。
在一种可能的实现方式中，在神经网络的训练过程中，可以将样本图像输入神经网络，得到样本图像中各类人脸属性的预测结果。例如，样本图像1包括3类人脸属性的标注信息，分别为人脸属性1的标注信息、人脸属性2的标注信息、人脸属性3的标注信息、人脸属性4的标注信息和人脸属性5的标 注信息。可以根据样本图像中人脸的实际属性，得到各人脸属性的标注信息。例如，当样本图像中的人脸不存在人脸属性3时，人脸属性3的标注信息可以为0或空。神经网络中预设的人脸属性包括5个，分别为人脸属性1至人脸属性5。例如，样本图像1针对各类人脸属性的预测结果为：人脸属性1的预测结果为80％、人脸属性2的预测结果为40％、人脸属性3的预测结果为60％、人脸属性4的预测结果为10％、人脸属性5的预测结果为20％。
在一种可能的实现方式中，可以根据各类人脸属性的预测结果，与各类人脸属性类人脸属性的标注信息，计算各类人脸属性的第一损失。可以根据预设的损失函数、目标人脸属性的预测结果和目标人脸属性的标注信息，计算目标人脸属性的第一损失。目标人脸属性为任一人脸属性。
在一种可能的实现方式中，可以将样本图像中各类人脸属性的第一损失相加得到第二损失。
在一种可能的实现方式中，步骤S300包括：根据目标人脸属性的第一损失和所述目标人脸属性的权值，确定所述目标人脸属性的权值损失，所述目标人脸属性包括任一所述人脸属性；根据各类所述人脸属性的权值损失，确定所述神经网络的第二损失。
在一种可能的实现方式中，可以为各类人脸属性设定不同的权值。例如，预设5类人脸属性，其中人脸属性1的权值为K1，人脸属性2的权值为K2……。可以将各类人脸属性的第一损失与其对应的权值相乘后，得到目标人脸属性的权值损失。再将各人脸属性的权值损失相加后，得到第二损失。
在一种可能的实现方式中，可以向神经网络反向传播第二损失的梯度，调整神经网络的网络参数。可以将样本图像集中的各样本图像依次输入神经网络，对神经网络进行迭代训练，直至达到预设的收敛条件或达到预设的训练次数，完成神经网络的训练。
在本实施例中，可以将样本图像输入神经网络得到各类人脸属性的预测结果，分别确定各人脸属性的预测结果与其标注信息的第一损失，根据各类人脸属性的第一损失确定第二损失，根据所述第二损失调整神经网络的网络参数。在本实施例中，根据各类人脸属性的损失确定的损失调整神经网络的参数。使得各类人脸属性之间的关联性，在神经网络的参数调整过程中得到充分的体现，可以提高神经网络进行多类人脸属性的识别的准确率。
图3示出根据本公开实施例的人脸属性识别方法中神经网络的训练流程示意图，如图3所示，所述神经网络的训练流程包括：
1、神经网络的卷积层提取人脸的特征。
2、神经网络的全连接层根据各预设的人脸属性，得到针对各类人脸属性的属性特征。如图3所示，得到人脸属性1的属性特征、人脸属性2的属性特征……人脸属性N的属性特征。
3、神经网络的分类层将各类人脸属性的属性特征进行分类处理，得到各类人脸属性的预测结果。根据各类人脸属性的预测结果和样本图像的标注信息，得到各类人脸属性的第一损失。
4、根据各类人脸属性的第一损失，和各类人脸属性对应的权值K，得到第二损失。
5、利用第二损失调整神经网络的参数，完成神经网络的训练。
在一种可能的实现方式中，所述神经网络的训练方法还包括：
对第一图像进行翻转、裁剪、平移、噪声添加中的至少一种，得到所述第一图像的噪声图像；将所述第一图像和所述噪声图像确定为样本图像，得到所述样本图像集。
在一种可能的实现方式中，可以用于训练神经网络的原始图像，进行水平翻转、随机剪裁、平移、颜色空间随机扰动等噪声添加，得到原始图像的噪声图像，并将噪声图像和原始图像均作为样本图像，得到样本图像集。
在本实施例中，通过对图像进行翻转、裁剪、平移、噪声添加中的至少一种，得到所述图像的噪声图像，可以增加样本图像的数量，提高神经网络的泛化能力。
在一种可能的实现方式中，步骤S10包括：调整所述图像的分辨率，得到第一处理图像；提取所述第一处理图像中人脸的特征。
在一种可能的实现方式中，可以通过调整图像的分辨率，将调整分辨率的图像输入神经网络进行人脸属性识别。例如可以根据需求，将图像的分辨率调整为224×224、128×128、96×96、64×64。图像的分辨率越低，神经网络进行人脸属性识别的速度越快。可以在满足人脸属性识别精度需求的前 提下，降低图像的分辨率以提高人脸属性识别的速率。
在本实施例中，可以通过调整图像的分辨率，提高神经网络进行人脸属性识别的速率。
在一种可能的实现方式中，所述方法还包括：
调整所述神经网络的特征的通道数，得到调整神经网络；
经神经网络的卷积层对所述图像进行卷积处理，以提取所述图像的卷积特征，包括：
经所述调整神经网络的卷积层对所述图像进行卷积处理，以提取所述图像的卷积特征。
在一种可能的实现方式中，可以调整神经网络的特征的通道数，包括调整神经网络中卷积层的卷积核的通道数。可以通过减少通道数，提高神经网络的人脸属性的识别速度。
图4示出根据本公开实施例的人脸属性识别装置的框图，如图4所示，所述人脸属性识别装置包括：
特征提取模块100，用于提取图像中人脸的特征；
概率预测结果获取模块200，用于根据所述人脸的特征进行预设的多类人脸属性的识别，得到各类人脸属性的概率预测结果；
属性识别结果获取模块300，用于根据各类人脸属性的概率预测结果确定所述图像中人脸的属性识别结果。
在一种可能的实现方式中，所述人脸的属性识别结果，包括；
所述人脸的至少两个属性识别结果。
在一种可能的实现方式中，所述人脸属性包括以下属性中的至少一种：年龄、性别、装饰物、肤色、表情、动作、种族。
在一种可能的实现方式中，所述装饰物包括普通眼镜、墨镜、帽子、口罩中的至少一种；所述表情包括：高兴、悲伤、哭泣、愤怒、激动中的至少一种；所述动作包括：抽烟、喝水/饮料、进食、化妆、打电话中的至少一种。
在一种可能的实现方式中，所述图像包括利用设置在车辆上的车载摄像头拍摄的车辆所在道路图像和/或车内图像，
所述图像中的人脸包括车辆所在道路上行人、其他车辆驾乘人员的人脸和/或车内驾乘人员的人脸，
所述属性识别结果，包括至少满足以下条件之一的属性识别结果：设定年龄、设定性别、设定装饰物、设定肤色、设定表情、设定动作、设定种族。
在一种可能的实现方式中，所述装置还包括：
第一目标对象确定模块，用于根据所述图像中人脸的属性识别结果确定第一目标对象；
驾驶控制模块，用于根据所述确定出的第一目标对象发送提示信息和/或向车辆控制系统发送控制指令，以使车辆执行以下驾驶动作之一：速度保持、方向保持、减速、转向、刹车。
在一种可能的实现方式中，所述图像包括利用设置在监控地点的监控摄像头拍摄的监控图像，所述监控地点包括以下任一地点：商场、学校、大型场馆、路面、办公楼宇；
所述属性识别结果，包括至少满足以下条件之一的属性识别结果：设定年龄、设定性别、设定装饰物、设定肤色、设定表情、设定动作、设定种族。
在一种可能的实现方式中，所述装置还包括：
第一目标对象确定模块，用于根据所述监控图像中人脸的属性识别结果确定第二目标对象；
监控控制模块，用于根据所述确定出的第二目标对象在所述监控地点发送设定信息，或
根据所述确定出的第二目标对象向监控人员发送提示信息。
在一种可能的实现方式中，所述特征提取模块，用于：
对所述图像进行卷积处理，以提取所述人脸的卷积特征。
在一种可能的实现方式中，所述概率预测结果获取模块，包括：
人脸属性特征提取子模块，用于提取所述人脸的卷积特征中各类人脸属性的特征，得到各类人脸属性的属性特征；
概率预测结果获取子模块，用于对所述各类人脸属性的属性特征进行分类处理，以确认各类人脸 属性的概率预测结果。
在一种可能的实现方式中，所述特征提取模块，用于：
经神经网络的卷积层对所述图像进行卷积处理，以提取所述图像的卷积特征；
所述人脸属性特征提取子模块，用于：
经所述神经网络的全连接层提取所述人脸的卷积特征中各类人脸属性的特征，得到各类人脸属性的属性特征；
所述概率预测结果获取子模块，用于：
经所述神经网络的分类层对所述各类人脸属性的属性特征进行分类处理，以确认各类人脸属性的概率预测结果。
在一种可能的实现方式中，所述神经网络基于样本图像集预先监督训练完成，其中，所述样本图像集中的样本图像包括有多类人脸属性的标注信息。
在一种可能的实现方式中，所述装置还包括训练模块，所述训练模块用于所述神经网络的训练，所述训练模块用于：
经神经网络分别获得所述样本图像中各类人脸属性的预测结果；
分别确定各类人脸属性的预测结果与其标注信息的第一损失；
根据各类人脸属性的第一损失确定第二损失；
根据所述第二损失调整神经网络的网络参数。
在一种可能的实现方式中，所述根据各类人脸属性的第一损失确定所述神经网络的第二损失，包括：
根据目标人脸属性的第一损失和所述目标人脸属性的权值，确定所述目标人脸属性的权值损失，所述目标人脸属性包括任一类所述人脸属性；
根据所述各类人脸属性的权值损失，确定所述神经网络的第二损失。
在一种可能的实现方式中，所述训练模块，还用于：对第一图像进行翻转、裁剪、平移、噪声添加中的至少一种，得到所述第一图像的噪声图像；
将所述第一图像和所述噪声图像确定为样本图像，得到所述样本图像集。
在一种可能的实现方式中，所述特征提取模块，包括：
第一处理图像获取模块，用于调整所述图像的分辨率，得到第一处理图像；
第一特征提取子模块，用于提取所述第一处理图像中人脸的特征。
在一种可能的实现方式中，所述装置还包括：
通道调整模块，用于调整所述神经网络的特征的通道数，得到调整神经网络；
所述特征提取模块，还用于：
经所述调整神经网络的卷积层对所述图像进行卷积处理，以提取所述图像的卷积特征。
可以理解，本公开提及的上述各个方法实施例，在不违背原理逻辑的情况下，均可以彼此相互结合形成结合后的实施例，限于篇幅，本公开不再赘述。
本领域技术人员可以理解，在具体实施方式的上述方法中，各步骤的撰写顺序并不意味着严格的执行顺序而对实施过程构成任何限定，各步骤的具体执行顺序应当以其功能和可能的内在逻辑确定。
在一些实施例中，本公开实施例提供的装置具有的功能或包含的模块可以用于执行上文方法实施例描述的方法，其具体实现可以参照上文方法实施例的描述，为了简洁，这里不再赘述
本公开实施例还提出一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述方法。计算机可读存储介质可以是非易失性计算机可读存储介质。
本公开实施例还提出一种电子设备，包括：处理器；用于存储处理器可执行指令的存储器；其中，所述处理器被配置为上述方法。
电子设备可以被提供为终端、服务器或其它形态的设备。
图5是根据一示例性实施例示出的一种电子设备800的框图。例如，电子设备800可以是移动电话， 计算机，数字广播终端，消息收发设备，游戏控制台，平板设备，医疗设备，健身设备，个人数字助理等终端。
参照图5，电子设备800可以包括以下一个或多个组件：处理组件802，存储器804，电源组件806，多媒体组件808，音频组件810，输入/输出(I/O)的接口812，传感器组件814，以及通信组件816。
处理组件802通常控制电子设备800的整体操作，诸如与显示，电话呼叫，数据通信，相机操作和记录操作相关联的操作。处理组件802可以包括一个或多个处理器820来执行指令，以完成上述的方法的全部或部分步骤。此外，处理组件802可以包括一个或多个模块，便于处理组件802和其他组件之间的交互。例如，处理组件802可以包括多媒体模块，以方便多媒体组件808和处理组件802之间的交互。
存储器804被配置为存储各种类型的数据以支持在电子设备800的操作。这些数据的示例包括用于在电子设备800上操作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图片，视频等。存储器804可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，如静态随机存取存储器(SRAM)，电可擦除可编程只读存储器(EEPROM)，可擦除可编程只读存储器(EPROM)，可编程只读存储器(PROM)，只读存储器(ROM)，磁存储器，快闪存储器，磁盘或光盘。
电源组件806为电子设备800的各种组件提供电力。电源组件806可以包括电源管理系统，一个或多个电源，及其他与为电子设备800生成、管理和分配电力相关联的组件。
多媒体组件808包括在所述电子设备800和用户之间的提供一个输出接口的屏幕。在一些实施例中，屏幕可以包括液晶显示器(LCD)和触摸面板(TP)。如果屏幕包括触摸面板，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括一个或多个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界，而且还检测与所述触摸或滑动操作相关的持续时间和压力。在一些实施例中，多媒体组件808包括一个前置摄像头和/或后置摄像头。当电子设备800处于操作模式，如拍摄模式或视频模式时，前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。
音频组件810被配置为输出和/或输入音频信号。例如，音频组件810包括一个麦克风(MIC)，当电子设备800处于操作模式，如呼叫模式、记录模式和语音识别模式时，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器804或经由通信组件816发送。在一些实施例中，音频组件810还包括一个扬声器，用于输出音频信号。
I/O接口812为处理组件802和外围接口模块之间提供接口，上述外围接口模块可以是键盘，点击轮，按钮等。这些按钮可包括但不限于：主页按钮、音量按钮、启动按钮和锁定按钮。
传感器组件814包括一个或多个传感器，用于为电子设备800提供各个方面的状态评估。例如，传感器组件814可以检测到电子设备800的打开/关闭状态，组件的相对定位，例如所述组件为电子设备800的显示器和小键盘，传感器组件814还可以检测电子设备800或电子设备800一个组件的位置改变，用户与电子设备800接触的存在或不存在，电子设备800方位或加速/减速和电子设备800的温度变化。传感器组件814可以包括接近传感器，被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件814还可以包括光传感器，如CMOS或CCD图像传感器，用于在成像应用中使用。在一些实施例中，该传感器组件814还可以包括加速度传感器，陀螺仪传感器，磁传感器，压力传感器或温度传感器。
通信组件816被配置为便于电子设备800和其他设备之间有线或无线方式的通信。电子设备800可以接入基于通信标准的无线网络，如WiFi，2G或3G，或它们的组合。在一个示例性实施例中，通信组件816经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中，所述通信组件816还包括近场通信(NFC)模块，以促进短程通信。例如，在NFC模块可基于射频识别(RFID)技术，红外数据协会(IrDA)技术，超宽带(UWB)技术，蓝牙(BT)技术和其他技术来实现。
在示例性实施例中，电子设备800可以被一个或多个应用专用集成电路(ASIC)、数字信号处理器(DSP)、数字信号处理设备(DSPD)、可编程逻辑器件(PLD)、现场可编程门阵列(FPGA)、控 制器、微控制器、微处理器或其他电子元件实现，用于执行上述方法。
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器804，上述计算机程序指令可由电子设备800的处理器820执行以完成上述方法。
本公开实施例还提供了一种计算机程序产品，包括计算机可读代码，当计算机可读代码在设备上运行时，设备中的处理器执行用于实现如上任一实施例提供的方法的指令。
该计算机程序产品可以具体通过硬件、软件或其结合的方式实现。在一个可选实施例中，所述计算机程序产品具体体现为计算机存储介质，在另一个可选实施例中，计算机程序产品具体体现为软件产品，例如软件开发包(Software Development Kit，SDK)等等。
图6是根据一示例性实施例示出的一种电子设备1900的框图。例如，电子设备1900可以被提供为一服务器。参照图6，电子设备1900包括处理组件1922，其进一步包括一个或多个处理器，以及由存储器1932所代表的存储器资源，用于存储可由处理组件1922的执行的指令，例如应用程序。存储器1932中存储的应用程序可以包括一个或一个以上的每一个对应于一组指令的模块。此外，处理组件1922被配置为执行指令，以执行上述方法。
电子设备1900还可以包括一个电源组件1926被配置为执行电子设备1900的电源管理，一个有线或无线网络接口1950被配置为将电子设备1900连接到网络，和一个输入输出(I/O)接口1958。电子设备1900可以操作基于存储在存储器1932的操作系统，例如Windows ServerTM，Mac OS XTM，UnixTM,LinuxTM，FreeBSDTM或类似。
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器1932，上述计算机程序指令可由电子设备1900的处理组件1922执行以完成上述方法。
本公开可以是系统、方法和/或计算机程序产品。计算机程序产品可以包括计算机可读存储介质，其上载有用于使处理器实现本公开的各个方面的计算机可读程序指令。
计算机可读存储介质可以是可以保持和存储由指令执行设备使用的指令的有形设备。计算机可读存储介质例如可以是――但不限于――电存储设备、磁存储设备、光存储设备、电磁存储设备、半导体存储设备或者上述的任意合适的组合。计算机可读存储介质的更具体的例子(非穷举的列表)包括：便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、静态随机存取存储器(SRAM)、便携式压缩盘只读存储器(CD-ROM)、数字多功能盘(DVD)、记忆棒、软盘、机械编码设备、例如其上存储有指令的打孔卡或凹槽内凸起结构、以及上述的任意合适的组合。这里所使用的计算机可读存储介质不被解释为瞬时信号本身，诸如无线电波或者其他自由传播的电磁波、通过波导或其他传输媒介传播的电磁波(例如，通过光纤电缆的光脉冲)、或者通过电线传输的电信号。
这里所描述的计算机可读程序指令可以从计算机可读存储介质下载到各个计算/处理设备，或者通过网络、例如因特网、局域网、广域网和/或无线网下载到外部计算机或外部存储设备。网络可以包括铜传输电缆、光纤传输、无线传输、路由器、防火墙、交换机、网关计算机和/或边缘服务器。每个计算/处理设备中的网络适配卡或者网络接口从网络接收计算机可读程序指令，并转发该计算机可读程序指令，以供存储在各个计算/处理设备中的计算机可读存储介质中。
用于执行本公开操作的计算机程序指令可以是汇编指令、指令集架构(ISA)指令、机器指令、机器相关指令、微代码、固件指令、状态设置数据、或者以一种或多种编程语言的任意组合编写的源代码或目标代码，所述编程语言包括面向对象的编程语言—诸如Smalltalk、C++等，以及常规的过程式编程语言—诸如“C”语言或类似的编程语言。计算机可读程序指令可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络—包括局域网(LAN)或广域网(WAN)—连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。在一些实施例中，通过利用计算机可读程序指令的状态信息来个性化定制电子电路，例如可编程逻辑电路、现场可编程门阵列(FPGA)或可编程逻辑阵列(PLA)，该电子电路可以执行计算机可读程序指令，从而实现本公开的各个方面。
这里参照根据本公开实施例的方法、装置(系统)和计算机程序产品的流程图和/或框图描述了本公开的各个方面。应当理解，流程图和/或框图的每个方框以及流程图和/或框图中各方框的组合，都可以由计算机可读程序指令实现。
这些计算机可读程序指令可以提供给通用计算机、专用计算机或其它可编程数据处理装置的处理器，从而生产出一种机器，使得这些指令在通过计算机或其它可编程数据处理装置的处理器执行时，产生了实现流程图和/或框图中的一个或多个方框中规定的功能/动作的装置。也可以把这些计算机可读程序指令存储在计算机可读存储介质中，这些指令使得计算机、可编程数据处理装置和/或其他设备以特定方式工作，从而，存储有指令的计算机可读介质则包括一个制造品，其包括实现流程图和/或框图中的一个或多个方框中规定的功能/动作的各个方面的指令。
也可以把计算机可读程序指令加载到计算机、其它可编程数据处理装置、或其它设备上，使得在计算机、其它可编程数据处理装置或其它设备上执行一系列操作步骤，以产生计算机实现的过程，从而使得在计算机、其它可编程数据处理装置、或其它设备上执行的指令实现流程图和/或框图中的一个或多个方框中规定的功能/动作。
附图中的流程图和框图显示了根据本公开的多个实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段或指令的一部分，所述模块、程序段或指令的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个连续的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或动作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
以上已经描述了本公开的各实施例，上述说明是示例性的，并非穷尽性的，并且也不限于所披露的各实施例。在不偏离所说明的各实施例的范围和精神的情况下，对于本技术领域的普通技术人员来说许多修改和变更都是显而易见的。本文中所用术语的选择，旨在最好地解释各实施例的原理、实际应用或对市场中的技术的技术改进，或者使本技术领域的其它普通技术人员能理解本文披露的各实施例。

Claims (37)

  1. 一种人脸属性识别方法，其特征在于，所述方法包括：
    提取图像中人脸的特征；
    根据所述人脸的特征进行预设的多类人脸属性的识别，得到各类人脸属性的概率预测结果；
    根据各类人脸属性的概率预测结果确定所述图像中人脸的属性识别结果。
  2. 根据权利要求1所述的方法，其特征在于，所述人脸的属性识别结果，包括：
    所述人脸的至少两个属性识别结果。
  3. 根据权利要求1或2所述的方法，其特征在于，所述人脸属性包括以下属性中的至少一种：年龄、性别、装饰物、肤色、表情、动作、种族。
  4. 根据权利要求3所述的方法，其特征在于，所述装饰物包括普通眼镜、墨镜、帽子、口罩中的至少一种；所述表情包括：高兴、悲伤、哭泣、愤怒、激动中的至少一种；所述动作包括：抽烟、喝水/饮料、进食、化妆、打电话中的至少一种。
  5. 根据权利要求3或4所述的方法，其特征在于，
    所述图像包括利用设置在车辆上的车载摄像头拍摄的车辆所在道路图像和/或车内图像，
    所述图像中的人脸包括车辆所在道路上行人、其他车辆驾乘人员的人脸和/或车内驾乘人员的人脸，
    所述属性识别结果，包括至少满足以下条件之一的属性识别结果：设定年龄、设定性别、设定装饰物、设定肤色、设定表情、设定动作、设定种族。
  6. 根据权利要求5所述的方法，其特征在于，所述方法还包括：
    根据所述图像中人脸的属性识别结果确定第一目标对象；
    根据所述确定出的第一目标对象发送提示信息和/或向车辆控制系统发送控制指令，以使车辆执行以下驾驶动作之一：速度保持、方向保持、减速、转向、刹车。
  7. 根据权利要求3或4所述的方法，其特征在于，
    所述图像包括利用设置在监控地点的监控摄像头拍摄的监控图像，所述监控地点包括以下任一地点：商场、学校、大型场馆、路面、办公楼宇；
    所述属性识别结果，包括至少满足以下条件之一的属性识别结果：设定年龄、设定性别、设定装饰物、设定肤色、设定表情、设定动作、设定种族。
  8. 根据权利要求7所述的方法，其特征在于，所述方法还包括：
    根据所述监控图像中人脸的属性识别结果确定第二目标对象；
    根据所述确定出的第二目标对象在所述监控地点发送设定信息，或
    根据所述确定出的第二目标对象向监控人员发送提示信息。
  9. 根据权利要求1至8中任一项所述的方法，其特征在于，所述提取图像中人脸的特征，包括：
    对所述图像进行卷积处理，以提取所述人脸的卷积特征。
  10. 根据权利要求9所述的方法，其特征在于，所述根据所述人脸的特征进行预设的多类人脸属性的识别，得到各类人脸属性的概率预测结果，包括：
    提取所述人脸的卷积特征中各类人脸属性的特征，得到各类人脸属性的属性特征；
    对所述各类人脸属性的属性特征进行分类处理，以确认各类人脸属性的概率预测结果。
  11. 根据权利要求10所述的方法，其特征在于，所述对所述图像进行卷积处理，以提取所述人脸的卷积特征，包括：
    经神经网络的卷积层对所述图像进行卷积处理，以提取所述图像的卷积特征；
    所述提取所述人脸的卷积特征中各类人脸属性的特征，得到各类人脸属性的属性特征，包括：
    经所述神经网络的全连接层提取所述人脸的卷积特征中各类人脸属性的特征，得到各类人脸属性的属性特征；
    所述对所述各类人脸属性的属性特征进行分类处理，以确认各类人脸属性的概率预测结果，包括：
    经所述神经网络的分类层对所述各类人脸属性的属性特征进行分类处理，以确认各类人脸属性的概率预测结果。
  12. 根据权利要求11所述的方法，其特征在于，所述神经网络基于样本图像集预先监督训练完成，其中，所述样本图像集中的样本图像包括有多类人脸属性的标注信息。
  13. 根据权利要求12所述的方法，其特征在于，所述神经网络的训练过程包括：
    经神经网络分别获得所述样本图像中各类人脸属性的预测结果；
    分别确定各类人脸属性的预测结果与其标注信息的第一损失；
    根据各类人脸属性的第一损失确定第二损失；
    根据所述第二损失调整神经网络的网络参数。
  14. 根据权利要求13所述的方法，其特征在于，所述根据各类人脸属性的第一损失确定所述神经网络的第二损失，包括：
    根据目标人脸属性的第一损失和所述目标人脸属性的权值，确定所述目标人脸属性的权值损失，所述目标人脸属性包括任一类所述人脸属性；
    根据所述各类人脸属性的权值损失，确定所述神经网络的第二损失。
  15. 根据权利要求12至14中任一项所述的方法，其特征在于，所述神经网络的训练方法还包括：
    对第一图像进行翻转、裁剪、平移、噪声添加中的至少一种，得到所述第一图像的噪声图像；
    将所述第一图像和所述噪声图像确定为样本图像，得到所述样本图像集。
  16. 根据权利要求1至15中任一项所述的方法，其特征在于，所述提取图像中人脸的特征，包括：
    调整所述图像的分辨率，得到第一处理图像；
    提取所述第一处理图像中人脸的特征。
  17. 根据权利要求11至16中任一项所述的方法，其特征在于，所述方法还包括：
    调整所述神经网络的特征的通道数，得到调整神经网络；
    经神经网络的卷积层对所述图像进行卷积处理，以提取所述图像的卷积特征，包括：
    经所述调整神经网络的卷积层对所述图像进行卷积处理，以提取所述图像的卷积特征。
  18. 一种人脸属性识别装置，其特征在于，所述装置包括：
    特征提取模块，用于提取图像中人脸的特征；
    概率预测结果获取模块，用于根据所述人脸的特征进行预设的多类人脸属性的识别，得到各类人脸属性的概率预测结果；
    属性识别结果获取模块，用于根据各类人脸属性的概率预测结果确定所述图像中人脸的属性识别结果。
  19. 根据权利要求18所述的装置，其特征在于，所述人脸的属性识别结果，包括：
    所述人脸的至少两个属性识别结果。
  20. 根据权利要求18或19所述的装置，其特征在于，所述人脸属性包括以下属性中的至少一种：年龄、性别、装饰物、肤色、表情、动作、种族。
  21. 根据权利要求20所述的装置，其特征在于，所述装饰物包括普通眼镜、墨镜、帽子、口罩中的至少一种；所述表情包括：高兴、悲伤、哭泣、愤怒、激动中的至少一种；所述动作包括：抽烟、喝水/饮料、进食、化妆、打电话中的至少一种。
  22. 根据权利要求20或21所述的装置，其特征在于，
    所述图像包括利用设置在车辆上的车载摄像头拍摄的车辆所在道路图像和/或车内图像，
    所述图像中的人脸包括车辆所在道路上行人、其他车辆驾乘人员的人脸和/或车内驾乘人员的人脸，
    所述属性识别结果，包括至少满足以下条件之一的属性识别结果：设定年龄、设定性别、设定装饰物、设定肤色、设定表情、设定动作、设定种族。
  23. 根据权利要求22所述的装置，其特征在于，所述装置还包括：
    第一目标对象确定模块，用于根据所述图像中人脸的属性识别结果确定第一目标对象；
    驾驶控制模块，用于根据所述确定出的第一目标对象发送提示信息和/或向车辆控制系统发送控制指令，以使车辆执行以下驾驶动作之一：速度保持、方向保持、减速、转向、刹车。
  24. 根据权利要求20或21所述的装置，其特征在于，
    所述图像包括利用设置在监控地点的监控摄像头拍摄的监控图像，所述监控地点包括以下任一地点：商场、学校、大型场馆、路面、办公楼宇；
    所述属性识别结果，包括至少满足以下条件之一的属性识别结果：设定年龄、设定性别、设定装饰物、设定肤色、设定表情、设定动作、设定种族。
  25. 根据权利要求24所述的装置，其特征在于，所述装置还包括：
    第一目标对象确定模块，用于根据所述监控图像中人脸的属性识别结果确定第二目标对象；
    监控控制模块，用于根据所述确定出的第二目标对象在所述监控地点发送设定信息，或
    根据所述确定出的第二目标对象向监控人员发送提示信息。
  26. 根据权利要求18至25中任一项所述的装置，其特征在于，所述特征提取模块，用于：
    对所述图像进行卷积处理，以提取所述人脸的卷积特征。
  27. 根据权利要求26所述的装置，其特征在于，所述概率预测结果获取模块，包括：
    人脸属性特征提取子模块，用于提取所述人脸的卷积特征中各类人脸属性的特征，得到各类人脸属性的属性特征；
    概率预测结果获取子模块，用于对所述各类人脸属性的属性特征进行分类处理，以确认各类人脸属性的概率预测结果。
  28. 根据权利要求27所述的装置，其特征在于，所述特征提取模块，用于：
    经神经网络的卷积层对所述图像进行卷积处理，以提取所述图像的卷积特征；
    所述人脸属性特征提取子模块，用于：
    经所述神经网络的全连接层提取所述人脸的卷积特征中各类人脸属性的特征，得到各类人脸属性的属性特征；
    所述概率预测结果获取子模块，用于：
    经所述神经网络的分类层对所述各类人脸属性的属性特征进行分类处理，以确认各类人脸属性的概率预测结果。
  29. 根据权利要求28所述的装置，其特征在于，所述神经网络基于样本图像集预先监督训练完成，其中，所述样本图像集中的样本图像包括有多类人脸属性的标注信息。
  30. 根据权利要求29所述的装置，其特征在于，所述装置还包括训练模块，所述训练模块用于所述神经网络的训练，所述训练模块用于：
    经神经网络分别获得所述样本图像中各类人脸属性的预测结果；
    分别确定各类人脸属性的预测结果与其标注信息的第一损失；
    根据各类人脸属性的第一损失确定第二损失；
    根据所述第二损失调整神经网络的网络参数。
  31. 根据权利要求30所述的装置，其特征在于，所述根据各类人脸属性的第一损失确定所述神经网络的第二损失，包括：
    根据目标人脸属性的第一损失和所述目标人脸属性的权值，确定所述目标人脸属性的权值损失，所述目标人脸属性包括任一类所述人脸属性；
    根据所述各类人脸属性的权值损失，确定所述神经网络的第二损失。
  32. 根据权利要求29至31中任一项所述的装置，其特征在于，所述训练模块，还用于：对第一图像进行翻转、裁剪、平移、噪声添加中的至少一种，得到所述第一图像的噪声图像；
    将所述第一图像和所述噪声图像确定为样本图像，得到所述样本图像集。
  33. 根据权利要求18至32中任一项所述的装置，其特征在于，所述特征提取模块，包括：
    第一处理图像获取模块，用于调整所述图像的分辨率，得到第一处理图像；
    第一特征提取子模块，用于提取所述第一处理图像中人脸的特征。
  34. 根据权利要求28至33中任一项所述的装置，其特征在于，所述装置还包括：
    通道调整模块，用于调整所述神经网络的特征的通道数，得到调整神经网络；
    所述特征提取模块，还用于：
    经所述调整神经网络的卷积层对所述图像进行卷积处理，以提取所述图像的卷积特征。
  35. 一种电子设备，其特征在于，包括：
    处理器；
    用于存储处理器可执行指令的存储器；
    其中，所述处理器被配置为：执行权利要求1至17中任意一项所述的方法。
  36. 一种计算机可读存储介质，其上存储有计算机程序指令，其特征在于，所述计算机程序指令被处理器执行时实现权利要求1至17中任意一项所述的方法。
  37. 一种计算机程序，包括计算机可读代码，当所述计算机可读代码在电子设备中运行时，所述电子设备中的处理器执行用于实现权利要求1-17中的任一权利要求所述的方法。
PCT/CN2019/121973 2018-12-29 2019-11-29 人脸属性识别方法及装置、电子设备和存储介质 Ceased WO2020134858A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
JP2021533534A JP2022515719A (ja) 2018-12-29 2019-11-29 顔属性認識方法及び装置、電子機器並びに記憶媒体
KR1020217020828A KR20210113607A (ko) 2018-12-29 2019-11-29 얼굴 속성 인식 방법 및 장치, 전자 기기 및 기억 매체

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811645459.1A CN111382642A (zh) 2018-12-29 2018-12-29 人脸属性识别方法及装置、电子设备和存储介质
CN201811645459.1 2018-12-29

Publications (1)

Publication Number Publication Date
WO2020134858A1 true WO2020134858A1 (zh) 2020-07-02

Family

ID=71126803

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/121973 Ceased WO2020134858A1 (zh) 2018-12-29 2019-11-29 人脸属性识别方法及装置、电子设备和存储介质

Country Status (4)

Country Link
JP (1) JP2022515719A (zh)
KR (1) KR20210113607A (zh)
CN (1) CN111382642A (zh)
WO (1) WO2020134858A1 (zh)

Cited By (24)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111860254A (zh) * 2020-07-10 2020-10-30 东莞正扬电子机械有限公司 一种驾驶员异常行为检测方法、装置、存储介质及设备
CN111860253A (zh) * 2020-07-10 2020-10-30 东莞正扬电子机械有限公司 一种驾驶场景的多任务属性识别方法、装置、介质及设备
CN112036369A (zh) * 2020-09-22 2020-12-04 北京中房智宝科技有限公司 一种人脸识别硬件装置
CN112232173A (zh) * 2020-10-12 2021-01-15 济南博观智能科技有限公司 一种行人属性识别方法、深度学习模型、设备及介质
CN112597886A (zh) * 2020-12-22 2021-04-02 成都商汤科技有限公司 乘车逃票检测方法及装置、电子设备和存储介质
CN112597944A (zh) * 2020-12-29 2021-04-02 北京市商汤科技开发有限公司 关键点检测方法及装置、电子设备和存储介质
CN112613447A (zh) * 2020-12-29 2021-04-06 上海商汤智能科技有限公司 关键点检测方法及装置、电子设备和存储介质
CN112651364A (zh) * 2020-12-31 2021-04-13 北京市商汤科技开发有限公司 图像处理方法、装置、电子设备及存储介质
CN113033374A (zh) * 2021-03-22 2021-06-25 开放智能机器（上海）有限公司 人工智能识别危险行为方法、装置、电子设备和存储介质
CN113221695A (zh) * 2021-04-29 2021-08-06 深圳数联天下智能科技有限公司 训练肤色识别模型的方法、识别肤色的方法及相关装置
CN113221086A (zh) * 2021-05-21 2021-08-06 深圳和锐网络科技有限公司 离线人脸认证方法、装置、电子设备及存储介质
CN113435390A (zh) * 2021-07-09 2021-09-24 上海商汤智能科技有限公司 人群定位方法及装置、电子设备和存储介质
CN113505674A (zh) * 2021-06-30 2021-10-15 上海商汤临港智能科技有限公司 人脸图像处理方法及装置、电子设备和存储介质
CN113762108A (zh) * 2021-08-23 2021-12-07 浙江大华技术股份有限公司 一种目标识别方法及装置
CN113920566A (zh) * 2021-10-29 2022-01-11 中国银联股份有限公司 图像处理方法、装置、电子设备及存储介质
CN114387633A (zh) * 2020-10-19 2022-04-22 北京澎思科技有限公司 面部属性识别方法、装置和计算机可读存储介质
CN114429611A (zh) * 2022-04-06 2022-05-03 北京达佳互联信息技术有限公司 视频合成方法、装置、电子设备及存储介质
CN114444594A (zh) * 2022-01-25 2022-05-06 阳光保险集团股份有限公司 一种吸烟识别的方法、装置、设备和可读存储介质
CN114862756A (zh) * 2022-03-25 2022-08-05 北京市商汤科技开发有限公司 对象测距方法及装置、电子设备和存储介质
WO2022205259A1 (zh) * 2021-04-01 2022-10-06 京东方科技集团股份有限公司 人脸属性检测方法及装置、存储介质及电子设备
CN115439815A (zh) * 2022-08-12 2022-12-06 北京罗克维尔斯科技有限公司 行驶工况识别方法、装置、设备、介质及车辆
WO2023178906A1 (zh) * 2022-03-22 2023-09-28 上海商汤智能科技有限公司 活体检测方法及装置、电子设备、存储介质、计算机程序、计算机程序产品
CN118942135A (zh) * 2024-07-15 2024-11-12 清华大学 高仿真人脸检测系统及方法、电子设备和存储介质
CN119540505A (zh) * 2024-10-28 2025-02-28 科利东方(北京)光电技术有限公司 电力杆塔金具识别控制方法和相关装置

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112699856A (zh) * 2021-03-24 2021-04-23 成都新希望金融信息有限公司 人脸装饰品识别方法、装置、电子设备及存储介质
CN113428176B (zh) * 2021-06-25 2023-11-14 阿波罗智联(北京)科技有限公司 无人车驾驶策略的调整方法、装置、设备和存储介质
CN113947803B (zh) * 2021-12-22 2022-03-25 北京的卢深视科技有限公司 模型训练、用于人脸识别的样本数据生成方法及电子设备
CN114332996B (zh) * 2021-12-22 2025-11-28 深圳市普渡科技有限公司 机器人播放显示内容的方法、装置、机器人以及介质
KR102769715B1 (ko) * 2022-07-20 2025-02-19 동국대학교 산학협력단 빛 번짐에 강인한 성별 인식 장치 및 방법

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106203395A (zh) * 2016-07-26 2016-12-07 厦门大学 基于多任务深度学习的人脸属性识别方法
CN106778682A (zh) * 2017-01-11 2017-05-31 厦门中控生物识别信息技术有限公司 一种卷积神经网络模型的训练方法及其设备
US20170193286A1 (en) * 2015-12-31 2017-07-06 Pinhole (Beijing) Technology Co., Ltd. Method and device for face recognition in video
CN108596011A (zh) * 2017-12-29 2018-09-28 中国电子科技集团公司信息科学研究院 一种基于组合深度网络的人脸属性识别方法和装置

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5100688B2 (ja) * 2009-03-10 2012-12-19 株式会社豊田中央研究所 対象物検出装置及びプログラム
EP2697742A4 (en) * 2011-04-11 2014-11-05 Intel Corp SYSTEM AND METHOD FOR SELECTING PERSONALIZED ADVERTISING
JP6254785B2 (ja) * 2012-07-24 2017-12-27 サイトセンシング株式会社 視聴率調査システム、並びに表情情報生成装置及び表情情報生成プログラム
EP3345127A4 (en) * 2015-08-31 2019-08-21 SRI International METHOD AND SYSTEM FOR MONITORING DRIVING BEHAVIOR
CN107784372B (zh) * 2016-08-24 2022-02-22 阿里巴巴集团控股有限公司 目标对象属性的预测方法、装置和系统
JP6788264B2 (ja) * 2016-09-29 2020-11-25 国立大学法人神戸大学 表情認識方法、表情認識装置、コンピュータプログラム及び広告管理システム
JP7169094B2 (ja) * 2017-06-01 2022-11-10 株式会社東芝 画像処理システム及び医用情報処理システム
CN107697069B (zh) * 2017-10-31 2020-07-28 上海汽车集团股份有限公司 汽车驾驶员疲劳驾驶智能控制方法
CN107844781A (zh) * 2017-11-28 2018-03-27 腾讯科技（深圳）有限公司 人脸属性识别方法及装置、电子设备及存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170193286A1 (en) * 2015-12-31 2017-07-06 Pinhole (Beijing) Technology Co., Ltd. Method and device for face recognition in video
CN106203395A (zh) * 2016-07-26 2016-12-07 厦门大学 基于多任务深度学习的人脸属性识别方法
CN106778682A (zh) * 2017-01-11 2017-05-31 厦门中控生物识别信息技术有限公司 一种卷积神经网络模型的训练方法及其设备
CN108596011A (zh) * 2017-12-29 2018-09-28 中国电子科技集团公司信息科学研究院 一种基于组合深度网络的人脸属性识别方法和装置

Cited By (31)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111860254A (zh) * 2020-07-10 2020-10-30 东莞正扬电子机械有限公司 一种驾驶员异常行为检测方法、装置、存储介质及设备
CN111860253A (zh) * 2020-07-10 2020-10-30 东莞正扬电子机械有限公司 一种驾驶场景的多任务属性识别方法、装置、介质及设备
CN111860253B (zh) * 2020-07-10 2024-11-08 广东正扬传感科技股份有限公司 一种驾驶场景的多任务属性识别方法、装置、介质及设备
CN112036369A (zh) * 2020-09-22 2020-12-04 北京中房智宝科技有限公司 一种人脸识别硬件装置
CN112232173A (zh) * 2020-10-12 2021-01-15 济南博观智能科技有限公司 一种行人属性识别方法、深度学习模型、设备及介质
CN112232173B (zh) * 2020-10-12 2023-04-07 济南博观智能科技有限公司 一种行人属性识别方法、深度学习模型、设备及介质
CN114387633A (zh) * 2020-10-19 2022-04-22 北京澎思科技有限公司 面部属性识别方法、装置和计算机可读存储介质
CN112597886A (zh) * 2020-12-22 2021-04-02 成都商汤科技有限公司 乘车逃票检测方法及装置、电子设备和存储介质
CN112597944A (zh) * 2020-12-29 2021-04-02 北京市商汤科技开发有限公司 关键点检测方法及装置、电子设备和存储介质
CN112597944B (zh) * 2020-12-29 2024-06-11 北京市商汤科技开发有限公司 关键点检测方法及装置、电子设备和存储介质
CN112613447A (zh) * 2020-12-29 2021-04-06 上海商汤智能科技有限公司 关键点检测方法及装置、电子设备和存储介质
CN112651364A (zh) * 2020-12-31 2021-04-13 北京市商汤科技开发有限公司 图像处理方法、装置、电子设备及存储介质
CN112651364B (zh) * 2020-12-31 2023-06-20 北京市商汤科技开发有限公司 图像处理方法、装置、电子设备及存储介质
CN113033374A (zh) * 2021-03-22 2021-06-25 开放智能机器（上海）有限公司 人工智能识别危险行为方法、装置、电子设备和存储介质
WO2022205259A1 (zh) * 2021-04-01 2022-10-06 京东方科技集团股份有限公司 人脸属性检测方法及装置、存储介质及电子设备
CN113221695B (zh) * 2021-04-29 2023-12-12 深圳数联天下智能科技有限公司 训练肤色识别模型的方法、识别肤色的方法及相关装置
CN113221695A (zh) * 2021-04-29 2021-08-06 深圳数联天下智能科技有限公司 训练肤色识别模型的方法、识别肤色的方法及相关装置
CN113221086A (zh) * 2021-05-21 2021-08-06 深圳和锐网络科技有限公司 离线人脸认证方法、装置、电子设备及存储介质
CN113505674A (zh) * 2021-06-30 2021-10-15 上海商汤临港智能科技有限公司 人脸图像处理方法及装置、电子设备和存储介质
CN113505674B (zh) * 2021-06-30 2023-04-18 上海商汤临港智能科技有限公司 人脸图像处理方法及装置、电子设备和存储介质
CN113435390A (zh) * 2021-07-09 2021-09-24 上海商汤智能科技有限公司 人群定位方法及装置、电子设备和存储介质
CN113762108A (zh) * 2021-08-23 2021-12-07 浙江大华技术股份有限公司 一种目标识别方法及装置
CN113920566A (zh) * 2021-10-29 2022-01-11 中国银联股份有限公司 图像处理方法、装置、电子设备及存储介质
CN114444594A (zh) * 2022-01-25 2022-05-06 阳光保险集团股份有限公司 一种吸烟识别的方法、装置、设备和可读存储介质
WO2023178906A1 (zh) * 2022-03-22 2023-09-28 上海商汤智能科技有限公司 活体检测方法及装置、电子设备、存储介质、计算机程序、计算机程序产品
CN114862756A (zh) * 2022-03-25 2022-08-05 北京市商汤科技开发有限公司 对象测距方法及装置、电子设备和存储介质
CN114429611B (zh) * 2022-04-06 2022-07-08 北京达佳互联信息技术有限公司 视频合成方法、装置、电子设备及存储介质
CN114429611A (zh) * 2022-04-06 2022-05-03 北京达佳互联信息技术有限公司 视频合成方法、装置、电子设备及存储介质
CN115439815A (zh) * 2022-08-12 2022-12-06 北京罗克维尔斯科技有限公司 行驶工况识别方法、装置、设备、介质及车辆
CN118942135A (zh) * 2024-07-15 2024-11-12 清华大学 高仿真人脸检测系统及方法、电子设备和存储介质
CN119540505A (zh) * 2024-10-28 2025-02-28 科利东方(北京)光电技术有限公司 电力杆塔金具识别控制方法和相关装置

Also Published As

Publication number Publication date
KR20210113607A (ko) 2021-09-16
CN111382642A (zh) 2020-07-07
JP2022515719A (ja) 2022-02-22

Similar Documents

Publication Publication Date Title
CN111382642A (zh) 人脸属性识别方法及装置、电子设备和存储介质
TWI759647B (zh) 影像處理方法、電子設備，和電腦可讀儲存介質
US10803323B2 (en) Electronic device and method of detecting driving event of vehicle
US20200042775A1 (en) Artificial intelligence server and method for de-identifying face area of unspecific person from image file
KR102060662B1 (ko) 차량의 주행 이벤트를 검출하는 전자 장치 및 방법
CN113411665B (zh) 提供视频信息的计算装置和方法
US10875525B2 (en) Ability enhancement
US20240297957A1 (en) Aspect ratio conversion for automated image generation
US20200151943A1 (en) Display system for presentation of augmented reality content
US20240169687A1 (en) Model training method, scene recognition method, and related device
CN120787344A (zh) 用于自动化的图像生成的提示修改
US20180189598A1 (en) Image Segmentation with Touch Interaction
WO2022227393A1 (zh) 图像拍摄方法及装置、电子设备和计算机可读存储介质
CN205720871U (zh) 一种智能抬头显示系统
WO2020220807A1 (zh) 图像生成方法及装置、电子设备及存储介质
WO2019212729A1 (en) Generating response based on user's profile and reasoning on contexts
KR102360181B1 (ko) 차량의 동작을 제어하는 전자 장치 및 방법
JP2022522551A (ja) 画像処理方法及び装置、電子機器並びに記憶媒体
KR20180109499A (ko) 사용자의 음성 입력에 대한 답변을 제공하는 방법 및 장치
KR20120120858A (ko) 영상통화 서비스 및 그 제공방법, 이를 위한 영상통화서비스 제공서버 및 제공단말기
CN111435422A (zh) 动作识别方法、控制方法及装置、电子设备和存储介质
CN115171678A (zh) 语音识别方法、装置、电子设备、存储介质及产品
CN115035596A (zh) 行为检测的方法及装置、电子设备和存储介质
CN110019960A (zh) 数据处理方法及装置、电子设备和存储介质
CN111541951A (zh) 基于视频的交互处理方法、装置、终端及可读存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19903843

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2021533534

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS (EPO FORM 1205A DATED 01.10.2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19903843

Country of ref document: EP

Kind code of ref document: A1