WO2020006964A1 - 图像检测方法和装置 - Google Patents

图像检测方法和装置 Download PDF

Info

Publication number
WO2020006964A1
WO2020006964A1 PCT/CN2018/116338 CN2018116338W WO2020006964A1 WO 2020006964 A1 WO2020006964 A1 WO 2020006964A1 CN 2018116338 W CN2018116338 W CN 2018116338W WO 2020006964 A1 WO2020006964 A1 WO 2020006964A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
preset
body part
detected
human body
Prior art date
Application number
PCT/CN2018/116338
Other languages
English (en)
French (fr)
Inventor
徐珍琦
朱延东
王长虎
Original Assignee
北京字节跳动网络技术有限公司
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 北京字节跳动网络技术有限公司 filed Critical 北京字节跳动网络技术有限公司
Publication of WO2020006964A1 publication Critical patent/WO2020006964A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • 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

Definitions

  • Embodiments of the present application relate to the field of computer technology, and in particular, to an image detection method and device.
  • the embodiments of the present application provide an image detection method and device.
  • an embodiment of the present application provides an image detection method.
  • the method includes: acquiring an image to be detected; performing preset body part recognition on the image to be detected to obtain at least one preset body part image included in the image to be detected; Position information; based on the obtained position information, a preset number of preset human body part images are intercepted from the image to be detected; the intercepted preset human body part images and the to-be-detected image are classified into preset category images to obtain classification result information; Based on the classification result information, detection result information used to characterize whether the image to be detected is a preset category image.
  • performing preset body part recognition on the image to be detected to obtain position information of at least one preset body part image included in the image to be detected includes: inputting the image to be detected into a pre-trained image for identifying presets.
  • a preset human body part recognition model of the human body part image obtains a recognition result, and the recognition result includes position information of at least one preset human body part image included in the image to be detected.
  • the recognition result further includes: category information and confidence of the human body part displayed in the at least one preset human body part image included in the image to be detected.
  • extracting a preset number of preset body part images from the image to be detected based on the obtained position information including: for at least one preset body part image, in the order of the confidence level from large to small, based on the obtained Position information from the to-be-detected image to capture a preset number of images of a preset body part.
  • classifying the intercepted preset human body part image and the image to be detected into preset category images to obtain classification result information includes: inputting the intercepted preset human body part image and the image to be detected into a pre-trained image.
  • an embodiment of the present application provides an image detection device, which includes: an acquisition unit configured to acquire an image to be detected; and an identification unit configured to perform preset body part recognition on the image to be detected to obtain the to-be-detected The position information of at least one preset human body part image included in the image; the interception unit is configured to intercept a preset number of preset human body part images from the image to be detected based on the obtained position information; the classification unit is configured to be respectively Classify the captured preset human body part image and the image to be detected into a preset category image to obtain classification result information; the generating unit is configured to generate a detection used to characterize whether the image to be detected is a preset category image based on the classification result information; Result information.
  • the recognition unit is further configured to: input the image to be detected into a pre-trained preset body part recognition model for recognizing a preset body part image to obtain a recognition result, and the recognition result includes the image to be detected includes Position information of at least one preset body part image.
  • the recognition result further includes: category information and confidence of the human body part displayed in the at least one preset human body part image included in the image to be detected.
  • the interception unit is further configured to: for at least one preset body part image, in a descending order of confidence, based on the obtained position information, intercept a preset number of presets from the image to be detected Body parts image.
  • the classification unit is further configured to: input the intercepted preset human body part image and the image to be detected into a pre-trained preset category image classification model for determining whether the image is a preset category image, and obtain Classification result information.
  • an embodiment of the present application provides an electronic device.
  • the electronic device includes: one or more processors; a storage device that stores one or more programs thereon; Or multiple processors execute, so that the above one or more processors implement the method as described in any implementation manner of the first aspect.
  • an embodiment of the present application provides a computer-readable medium on which a computer program is stored.
  • a computer program is stored.
  • the image detection method and device provided in the embodiments of the present application first perform recognition of a preset human body part on an image to be detected, and obtain position information of a preset human body part image included in the image to be detected. Therefore, based on the obtained position information, a preset number of preset human body part images can be intercepted from the image to be detected. After that, the classified images of the preset human body parts and the images to be detected are classified into preset category images to obtain classification result information. Finally, based on the classification result information, detection result information used to characterize whether the image to be detected is an image of a preset category is generated. Because the detection result information is generated based on the image to be detected and the intercepted preset human body part image. Therefore, the detection result information combines the overall information and local information of the image to be detected, thereby improving the accuracy of the detection result information.
  • FIG. 1 is an exemplary system architecture diagram to which an embodiment of the present application can be applied;
  • FIG. 1 is an exemplary system architecture diagram to which an embodiment of the present application can be applied;
  • FIG. 2 is a flowchart of an embodiment of an image detection method according to the present application.
  • FIG. 3 is a schematic diagram of an application scenario of the image detection method according to the present application.
  • FIG. 4 is a flowchart of another embodiment of an image detection method according to the present application.
  • FIG. 5 is a schematic structural diagram of an embodiment of an image detection device according to the present application.
  • FIG. 6 is a schematic structural diagram of a computer system suitable for implementing an electronic device according to an embodiment of the present application.
  • FIG. 1 illustrates an exemplary system architecture 100 to which an image detection method or an image detection apparatus of an embodiment of the present application can be applied.
  • the system architecture 100 may include terminals 101, 102, and 103, a network 104, and a server 105.
  • the network 104 is a medium for providing a communication link between the terminal devices 101, 102, 103 and the server 105.
  • the network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.
  • the terminal devices 101, 102, and 103 interact with the server 105 through the network 104, for example, send an image to be detected to the server 105.
  • Various types of camera applications, picture processing applications, etc. can be installed on the terminal devices 101, 102, 103.
  • the terminal devices 101, 102, and 103 may be hardware or software.
  • the terminal devices 101, 102, and 103 may be devices capable of capturing or storing images, including but not limited to: cameras, mobile phones with photographing functions, picture storage servers, and the like.
  • the terminal devices 101, 102, and 103 are software, they can be installed in the electronic devices listed above. It can be implemented as multiple software or software modules (for example, to provide photographing services), or it can be implemented as a single software or software module. It is not specifically limited here.
  • the server 105 may be a server providing various services, for example, detecting the images to be detected acquired from the terminal devices 101, 102, and 103 to obtain detection result information. If necessary, the detection result information may also be sent to the terminal devices 101, 102, and 103.
  • the image detection method provided in the embodiment of the present application may be executed by the server 105 or a terminal device. Accordingly, the image detection device may be provided in the server 105 or in a terminal device.
  • the terminal devices 101, 102, and 103 may also detect images.
  • the image detection method may also be executed by the terminal devices 101, 102, and 103.
  • the image detection device may also be provided in the terminal devices 101, 102, 103.
  • the exemplary system architecture 100 may be absent from the server 105 and the network 104.
  • the server may be hardware or software.
  • the server can be implemented as a distributed server cluster consisting of multiple servers or as a single server.
  • the server can be implemented as multiple software or software modules (for example, to provide distributed services), or it can be implemented as a single software or software module. It is not specifically limited here.
  • terminal devices, networks, and servers in FIG. 1 are merely exemplary. Depending on the implementation needs, there can be any number of terminal devices, networks, and servers.
  • the image detection method includes the following steps:
  • Step 201 Acquire an image to be detected.
  • the execution subject of the image detection method may acquire the image to be detected from the terminal device through a wired connection method or a wireless connection method.
  • the image to be detected may be any image.
  • the determination of the image to be detected can be specified by a technician, or it can be filtered according to certain conditions.
  • the image to be detected may also be stored locally on the execution subject. At this time, the execution subject may directly obtain the image to be detected locally.
  • Step 202 Recognize a preset body part image to be detected, and obtain position information of at least one preset body part image included in the image to be detected.
  • the execution subject may perform preset body part recognition by using various methods on the image to be detected to obtain position information of at least one preset body part image included in the image to be detected.
  • the preset human body part may be at least one part of the human body, including but not limited to at least one of the following: mouth, eyes, nose, female chest, female reproductive organ, male reproductive organ, and the like.
  • the position information is used to represent a position of a preset image of a human body part relative to an image to be detected. Position information comes in various forms, such as callout boxes, coordinates, and so on.
  • the above-mentioned execution subject may perform preset body part recognition through a cascade classifier, thereby obtaining position information of at least one preset body part image included in an image to be detected.
  • the cascaded classifier may be a cascaded classifier (such as a Haar classifier).
  • OpenCV Open Source Computer Vision Library
  • the detection rate may be a probability of detecting a preset body part displayed in the image.
  • Multiple classifiers can be cascaded in order of detection rate.
  • the execution body may input the image to be detected into a cascade classifier.
  • the classifier with the highest detection rate is used to detect the image to be detected. If the image to be detected displays a preset human body part, the image to be detected is sent to the next classifier. Subsequent classifiers detect in the same way, until the one with the lowest detection rate. If it is detected that the preset human body part is not displayed in the image, the information that the preset human body part is not displayed in the image can be characterized. If the last-level classifier (the classifier with the lowest detection rate) detects that a preset human body part is displayed in the image, it can output a label box used to indicate the position information of the preset human body part displayed in the image. Thereby, position information of at least one preset human body part image included in the image to be detected is obtained.
  • an image to be detected may be input into a pre-trained preset body part recognition model for identifying a preset body part image to obtain a recognition result, and the recognition result includes the image to be detected. Position information of at least one preset human body part image is included.
  • the preset human body part recognition model can be used to detect whether the image contains a preset human body part image, that is, whether the preset human body part is displayed in the image.
  • the preset human body part recognition model can be trained by the following steps:
  • the executing agent can obtain a training sample set.
  • Each training sample in the training sample set may include a sample image and label information of the sample image.
  • the annotation information includes position information of an image of a preset human body part included in the sample image and category information of the preset human body part displayed in the sample image.
  • the position information is used to represent a position of an image of a preset human body part relative to a sample image. Position information comes in various forms, such as callout boxes, coordinates, and so on.
  • the category information of the preset human body part is used to indicate the category of the preset human body part.
  • the preset human body part may be at least one human body part.
  • the preset human body part may include three parts of a female chest, a male genitalia, and a female genitalia. Then, the category information of the preset human body part may be “00”, “01”, and “10”, which are respectively used to indicate these three parts.
  • the executing body may take as input the sample images of the training samples in the training sample set, and use the label information corresponding to the input sample images as the desired output, and train to obtain a preset human body part detection model.
  • the sample images of the training samples in the training sample set may be input into an initial preset human body part detection model.
  • the initial preset human body part detection model may be various target detection networks. As an example, it can be an existing SSD (Single Shot MultiBox Detector) or YOLO (You Only Look Out). In practice, an initial value can be set for an initial preset human body part detection model. For example, it could be some different small random numbers.
  • the "small random number” is used to ensure that the network does not enter a saturation state due to excessive weights, which causes training failure. "Different” is used to ensure that the network can learn normally.
  • the detection result of the input sample image can be obtained.
  • the machine learning method is used to train the initial preset human body part detection model. Specifically, the difference between the detection result and the labeled information calculated by using a preset loss function may be used first.
  • the parameters of the initial preset human body part detection model can be adjusted, and if the preset training end conditions are satisfied, the training is ended, and the trained initial preset human body part detection model is used as a pre- Set up a human body part detection model.
  • the training end condition here includes but is not limited to at least one of the following: the training time exceeds a preset duration; the number of training times reaches a preset number of times; and the calculated difference is less than a preset difference threshold.
  • BP Back Propagation, Back Propagation
  • SGD Spochastic Gradient Descent, Stochastic Gradient Descent
  • the execution subject of the training step and the execution subject of the image detection method may be the same or different. If they are the same, the executing subject may store the network structure and parameter values of the preset human body part detection model locally after training to obtain the preset human body part detection model. If they are different, after the training subject has obtained a preset human body part detection model through training, the network structure and parameter values of the model can be sent to the image detection method execution subject.
  • step 203 based on the obtained position information, a preset number of preset human body part images are intercepted from the image to be detected.
  • the execution subject uses various methods (for example, various screenshot applications) to intercept a preset number of preset human body part images from the image to be detected.
  • the position information may be a rectangular callout box.
  • rectangular callout boxes can be represented by vectors. The vector can include the coordinates of the geometric center of the rectangular callout box, and the height and width of the rectangular callout box.
  • Step 204 Classify the images of the preset human body parts and the images to be detected into preset category images to obtain classification result information.
  • the above-mentioned execution subject may use various methods to perform classification of the preset category image on the intercepted preset human body part image and the image to be detected, to obtain classification result information.
  • the preset category images may be images of various categories. For example, it may be a facial image, a vulgar erotic image, and the like. It should be noted that the preset category image here matches the preset human body part image. Specifically, if the preset category image is a face image. Then, the preset body part image may be a mouth image, an eye image, and the like. If the preset category image is vulgar erotic image. Then, the preset body part image may be a chest image, a human reproductive organ image, and the like.
  • the classification result information is used to characterize whether the image is a preset category image.
  • the classification result information may be various forms of information.
  • the classification result information may be a numerical value. For example, "0" indicates that it is not a preset category image, and "1" indicates that it is a preset category image.
  • the classification result information may also be text, characters, and so on.
  • the above-mentioned execution subject may use a classification model (such as a word bag model) to perform classification of the preset category image on the intercepted preset human body part image and the image to be detected, to obtain classification result information.
  • a classification model such as a word bag model
  • the word bag model should be widely used in image recognition, and its implementation can include feature extraction, feature encoding, feature aggregation, and classification using a classifier.
  • feature extraction can use various detection operators, such as Harris corner detection operator, FAST (Features from Accelerated Segment Test) operator, etc., to perform feature extraction on objects.
  • features can be encoded. For example, querying a dictionary to achieve feature encoding. After that, the encoded multiple features are stitched and used as the final expression of the image.
  • the specific expression form can be a vector.
  • classifiers such as support vector machines are used to classify the resulting vectors.
  • the classification result information is obtained.
  • Step 205 Based on the classification result information, generate detection result information used to characterize whether the image to be detected is an image of a preset category.
  • the above-mentioned execution subject may use various methods to generate detection result information used to characterize whether the image to be detected is a preset category image based on the classification result information.
  • the classification result information obtained in step 204 may include first classification result information and second classification result information.
  • the first classification result information is used to characterize whether the captured preset human body part image is a preset category image.
  • the second classification result information is used to characterize whether the image to be detected is a preset category image.
  • the execution body may input the first classification result information and the second classification result information into the classifier.
  • the classifier may be a Softmax classifier. Softmax can map input information to the (0,1) interval. And all input information will be used in the calculation process. After that, the output of the classifier corresponding to the second classification result information may be used as the detection result information.
  • the execution entity may also obtain the detection result information by querying a preset correspondence table between classification result information and detection result information.
  • the correspondence relationship table may be obtained based on a large number of statistics.
  • Correspondence characterization can record a large amount of result information and corresponding detection result information.
  • the execution body may query the classification result information obtained in step 204 in a correspondence relationship table. If there is classification result information matching the obtained classification result information, it may obtain detection result information corresponding to the matching classification result information. After that, the obtained detection result information can be used as the detection result information used to characterize whether the image to be detected is an image of a preset category.
  • FIG. 3 is a schematic diagram of an application scenario of the image detection method according to this embodiment.
  • the execution subject of the image detection method may be the server 300.
  • the server 300 first obtains an image 301 to be detected. After that, preset human body part recognition is performed on the image to be detected.
  • the preset body part may be a chest.
  • the position information of the chest image included in the to-be-detected image 301 is obtained, and is shown as a callout box 302 in the figure.
  • the server 300 captures a chest image from the image 301 to be detected, and obtains a captured image 303.
  • the captured image 303 and the to-be-detected image 301 are respectively input into a classification model (for example, a word bag model) 304 to obtain classification result information.
  • the classification result information of the captured image 303 and the to-be-detected image 301 are both "1". It is indicated that the captured image 303 and the to-be-detected image 301 are both preset category images (for example, bad images).
  • the server 300 may query the correspondence table 305. By querying, the classification result information is "1" and the detection result information is also "1". Thereby, detection result information for characterizing whether the image to be detected is a preset category image (for example, a bad image) can be obtained.
  • the method provided by the foregoing embodiment of the present application may first acquire an image to be detected. After that, preset human body part recognition is performed on the image to be detected to obtain position information of at least one preset human body part image included in the image to be detected. Therefore, based on the obtained position information, a preset number of preset human body part images can be intercepted from the image to be detected. Then, classification of the preset image of the human body part and the image to be detected are performed for the preset category image to obtain classification result information. Finally, based on the classification result information, detection result information that is used to characterize whether the image to be detected is a preset category image is generated. Because the detection result information is generated based on the image to be detected and the intercepted preset human body part image. Therefore, the detection result information combines the overall information and local information of the image to be detected, thereby improving the accuracy of the detection result information.
  • FIG. 4 illustrates a process 400 of still another embodiment of the image detection method.
  • the process 400 of the image detection method includes the following steps:
  • Step 401 Acquire an image to be detected.
  • step 401 is similar to step 201 of the embodiment corresponding to FIG. 2, and details are not described herein again.
  • Step 402 Input a to-be-detected image into a pre-trained preset human body part recognition model for identifying a preset human body part image to obtain a recognition result.
  • the recognition result includes position information of at least one preset human body part image included in the image to be detected, category information and confidence of the human body part displayed in the at least one preset human body part image included in the image to be detected.
  • the confidence degree is used to indicate the credibility of the category information.
  • confidence can be expressed as a probability value.
  • the execution subject of the image detection method may input the image to be detected into a preset human body part recognition model that is pre-trained and used to identify a preset human body part image.
  • the preset human body part recognition model is used to recognize a preset human body part image in the image.
  • step 403 a preset number of preset human body part images are intercepted from the image to be detected based on the obtained position information in order of increasing confidence.
  • the image to be detected may include at least one preset human body part image. Therefore, the above-mentioned execution subject may intercept a preset number of preset human body part images from the image to be detected based on the obtained position information in the descending order of confidence.
  • step 404 the captured preset human body part image and the image to be detected are input into a preset category image classification model that is pre-trained and used to determine whether the image is a preset category image, to obtain classification result information.
  • the above-mentioned execution subject may respectively input the intercepted preset human body part image and the image to be detected into a pre-trained preset category image classification model to obtain classification result information.
  • the classification result information may be information used to characterize whether the image is a preset category image.
  • the classification result information herein may include first classification result information and second classification result information.
  • the first classification result information may be used to characterize whether the captured preset human body part image is a preset category image.
  • the second classification result information may be used to characterize whether the image to be detected is a preset category image.
  • the preset category image may be an image of any category. As examples, it may be a face image, a head image, a vulgar erotic image, and the like.
  • the preset category image classification model is used to determine whether an image is a preset category image.
  • the preset category image classification model can be trained by the following steps:
  • the first step is to obtain a training sample set.
  • Each training sample in the training sample set includes a sample image and label information.
  • the labeling information is used to represent whether the sample image is a preset category image.
  • the label information may be in various forms.
  • the label information may be a numerical value. For example, "0" indicates that it is not a preset category image, and "1" indicates that it is a preset category image.
  • the label information may also be text, characters, and so on.
  • the sample image of the training sample is used as input, and the label information corresponding to the input sample image is used as the desired output.
  • the machine learning method is used to train and obtain a preset category classification model .
  • the initial preset category image classification model can be trained based on the training sample set to obtain the preset category image classification model.
  • the initial preset category image classification model may be various image classification networks. As an example, it can be a residual network (Residual Network, ResNet), VGG, or the like.
  • VGG is a classification model proposed by the Visual Geometry Group (VGG) of a university.
  • a sample image of the training sample may be input into an initial preset category classification model.
  • an initial value can be set for the initial preset category classification model.
  • the classification result information of the input sample image can be obtained.
  • a difference between the classification result information and the label information calculated by a preset loss function can be used.
  • the parameters of the initial preset category classification model can be adjusted based on the obtained differences, and if the preset training end condition is met, the training is ended, and the trained initial preset category classification model is used as the preset category image classification. model.
  • the training end condition here includes but is not limited to at least one of the following: the training time exceeds a preset duration; the number of training times reaches a preset number of times; and the calculated difference is less than a preset difference threshold.
  • Step 405 Based on the classification result information, generate detection result information used to characterize whether the image to be detected is an image of a preset category.
  • step 405 the specific implementation of step 405 and the technical effects that it brings are similar to step 205 in the embodiment corresponding to FIG. 2, and details are not described herein again.
  • the image detection method in this embodiment obtains a recognition result by inputting an image to be detected into a preset human body part recognition model.
  • a preset number of images of a preset body part are captured.
  • the accuracy of preset human body part recognition is improved. Thereby, the detection accuracy of the image to be detected is improved.
  • this application provides an embodiment of an image detection device.
  • the device embodiment corresponds to the method embodiment shown in FIG. 2, and the device may specifically Used in various electronic equipment.
  • the image detection device 500 in this embodiment includes an acquisition unit 501, a recognition unit 502, a cropping unit 503, a classification unit 504, and a generation unit 505.
  • the obtaining unit 501 is configured to obtain an image to be detected.
  • the recognition unit 502 is configured to perform preset body part recognition on the image to be detected, and obtain position information of at least one preset body part image included in the image to be detected.
  • the cropping unit 503 is configured to crop a preset number of preset human body part images from the image to be detected based on the obtained position information.
  • the classification unit 504 is configured to classify the intercepted image of the preset human body part and the image to be detected in a preset category to obtain classification result information.
  • the generating unit 505 is configured to generate detection result information used to characterize whether the image to be detected is a preset category image based on the classification result information.
  • the recognition unit is further configured to: input the image to be detected into a pre-trained preset human body part recognition model for identifying a preset human body part image, obtain a recognition result, and recognize The result includes position information of at least one preset human body part image included in the image to be detected.
  • the recognition result may further include: category information and confidence of the human body part displayed in the at least one preset human body part image included in the image to be detected.
  • the interception unit 503 may be further configured to: for at least one preset image of a human body part, in order of the confidence level from large to small, based on the obtained position information, from the to-be-detected A predetermined number of images of a preset body part are captured in the image.
  • the classification unit 504 may be further configured to: input the intercepted preset human body part image and the image to be detected into a pre-trained image for determining whether the image is a preset category image Image classification model of a preset category to obtain classification result information.
  • the image detection apparatus may first obtain an image to be detected through the acquisition unit 501.
  • the recognition unit 502 recognizes the preset human body part image to obtain the position information of at least one preset human body part image included in the image to be detected. Therefore, the intercepting unit 503 can intercept a preset number of preset human body part images from the image to be detected based on the obtained position information.
  • the classification unit 504 performs classification of the preset category images on the intercepted images of the preset human body parts and the images to be detected, to obtain classification result information.
  • the generating unit 505 generates detection result information used to characterize whether the image to be detected is a preset category image based on the classification result information. Because the detection result information is generated based on the image to be detected and the intercepted preset human body part image. Therefore, the detection result information combines the overall information and local information of the image to be detected, thereby improving the accuracy of the detection result information.
  • FIG. 6 a schematic structural diagram of a computer system 600 suitable for implementing an electronic device according to an embodiment of the present application is shown.
  • the electronic device shown in FIG. 6 is only an example, and should not impose any limitation on the functions and scope of use of the embodiments of the present application.
  • the computer system 600 includes a central processing unit (CPU) 601, which can be loaded into a random access memory (RAM) 603 from a program stored in a read-only memory (ROM) 602 or from a storage portion 608 Instead, perform various appropriate actions and processes.
  • RAM random access memory
  • ROM read-only memory
  • various programs and data required for the operation of the system 600 are also stored.
  • the CPU 601, the ROM 602, and the RAM 603 are connected to each other through a bus 604.
  • An input / output (I / O) interface 605 is also connected to the bus 604.
  • the following components are connected to the I / O interface 605: an input portion 606 including a keyboard, a mouse, and the like; an output portion 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage portion 608 including a hard disk and the like; a communication section 609 including a network interface card such as a LAN card, a modem, and the like.
  • the communication section 609 performs communication processing via a network such as the Internet.
  • the driver 610 is also connected to the I / O interface 605 as necessary.
  • a removable medium 611 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed into the storage section 608 as needed.
  • the process described above with reference to the flowchart may be implemented as a computer software program.
  • embodiments of the present disclosure include a computer program product including a computer program carried on a computer-readable medium, the computer program containing program code for performing a method shown in a flowchart.
  • the computer program may be downloaded and installed from a network through the communication section 609, and / or installed from a removable medium 611.
  • CPU central processing unit
  • the computer-readable medium described in this application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two.
  • the computer-readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable Programming read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the foregoing.
  • a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in combination with an instruction execution system, apparatus, or device.
  • a computer-readable signal medium may include a data signal that is included in baseband or propagated as part of a carrier wave, and which carries computer-readable program code. Such a propagated data signal may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing.
  • the computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, and the computer-readable medium may send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device .
  • Program code embodied on a computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
  • Computer program code for performing the operations of this application may be written in one or more programming languages, or a combination thereof, including programming languages such as Java, Smalltalk, C ++, and also conventional Procedural programming language—such as "C" or a similar programming language.
  • the program code 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, partly on a remote computer, or entirely on a remote computer or server.
  • the remote computer can 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 it can be connected to an external computer (such as through an Internet service provider) Internet connection).
  • LAN local area network
  • WAN wide area network
  • Internet service provider Internet service provider
  • each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, which contains one or more functions to implement a specified logical function Executable instructions.
  • the functions labeled in the blocks may also occur in a different order than those labeled in the drawings. For example, two blocks represented one after the other may actually be executed substantially in parallel, and they may sometimes be executed in the 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 by a dedicated hardware-based system that performs the specified function or operation , Or it can be implemented with a combination of dedicated hardware and computer instructions.
  • the units described in the embodiments of the present application may be implemented by software or hardware.
  • the described unit may also be provided in a processor, for example, it may be described as: a processor includes an acquisition unit, a recognition unit, an interception unit, a classification unit, and a generation unit.
  • a processor includes an acquisition unit, a recognition unit, an interception unit, a classification unit, and a generation unit.
  • the names of these units do not constitute a limitation on the unit itself in some cases.
  • the obtaining unit may also be described as a “unit for obtaining an image to be detected”.
  • the present application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments; or may exist separately without being assembled into the electronic device in.
  • the computer-readable medium carries one or more programs, and when the one or more programs are executed by the electronic device, the electronic device: obtains an image to be detected; performs preset body part recognition on the image to be detected, and obtains the to-be-detected Position information of at least one preset human body part image included in the image; based on the obtained position information, a preset number of preset human body part images are intercepted from the image to be detected; the intercepted preset human body part image and the image to be detected are respectively Perform classification of the preset category image to obtain classification result information; and based on the classification result information, generate detection result information used to characterize whether the image to be detected is a preset category image.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

本申请实施例公开了图像检测方法和装置。该方法的一具体实施方式包括：获取待检测图像；对待检测图像进行预设人体部位识别，得到待检测图像中包括的至少一个预设人体部位图像的位置信息；基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像；分别对截取的预设人体部位图像和待检测图像进行预设类别图像分类，得到分类结果信息；基于分类结果信息，生成用于表征待检测图像是否为预设类别图像的检测结果信息。该实施方式实现了提高了检测结果信息的准确率。

Description

图像检测方法和装置
本专利申请要求于2018年7月6日提交的、申请号为201810734680.8、申请人为北京字节跳动网络技术有限公司、发明名称为“图像检测方法和装置”的中国专利申请的优先权，该申请的全文以引用的方式并入本申请中。
技术领域
本申请实施例涉及计算机技术领域，具体涉及图像检测方法和装置。
背景技术
随着互联网的快速发展，尤其是移动互联网的普及，各种内容的视频或图像层出不穷。目前，主要采取人工审核的方式对这些视频或图像的内容进行审核。
发明内容
本申请实施例提出了图像检测方法和装置。
第一方面，本申请实施例提供了一种图像检测方法，该方法包括：获取待检测图像；对待检测图像进行预设人体部位识别，得到待检测图像中包括的至少一个预设人体部位图像的位置信息；基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像；分别对截取的预设人体部位图像和待检测图像进行预设类别图像分类，得到分类结果信息；基于分类结果信息，生成用于表征待检测图像是否为预设类别图像的检测结果信息。
在一些实施例中，对待检测图像进行预设人体部位识别，得到待检测图像中包括的至少一个预设人体部位图像的位置信息，包括：将待检测图像输入预先训练的、用于识别预设人体部位图像的预设人体 部位识别模型，得到识别结果，识别结果包括待检测图像中包括的至少一个预设人体部位图像的位置信息。
在一些实施例中，识别结果还包括：待检测图像中包括的至少一个预设人体部位图像所显示的人体部位的类别信息和置信度。
在一些实施例中，基于得到的位置信息，从待检测图像截取预设数量的预设人体部位图像，包括：对于至少一个预设人体部位图像，按照置信度由大到小的顺序，基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像。
在一些实施例中，分别对截取的预设人体部位图像和待检测图像进行预设类别图像分类，得到分类结果信息，包括：分别将截取的预设人体部位图像和待检测图像输入预先训练的、用于确定图像是否为预设类别图像的预设类别图像分类模型，得到分类结果信息。
第二方面，本申请实施例提供了一种图像检测装置，该装置包括：获取单元，被配置成获取待检测图像；识别单元，被配置成对待检测图像进行预设人体部位识别，得到待检测图像中包括的至少一个预设人体部位图像的位置信息；截取单元，被配置成基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像；分类单元，被配置成分别对截取的预设人体部位图像和待检测图像进行预设类别图像分类，得到分类结果信息；生成单元，被配置成基于分类结果信息，生成用于表征待检测图像是否为预设类别图像的检测结果信息。
在一些实施例中，识别单元进一步被配置成：将待检测图像输入预先训练的、用于识别预设人体部位图像的预设人体部位识别模型，得到识别结果，识别结果包括待检测图像中包括的至少一个预设人体部位图像的位置信息。
在一些实施例中，识别结果还包括：待检测图像中包括的至少一个预设人体部位图像所显示的人体部位的类别信息和置信度。
在一些实施例中，截取单元进一步被配置成：对于至少一个预设人体部位图像，按照置信度由大到小的顺序，基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像。
在一些实施例中，分类单元进一步被配置成：分别将截取的预设 人体部位图像和待检测图像输入预先训练的、用于确定图像是否为预设类别图像的预设类别图像分类模型，得到分类结果信息。
第三方面，本申请实施例提供了一种电子设备，该电子设备包括：一个或多个处理器；存储装置，其上存储有一个或多个程序；当上述一个或多个程序被上述一个或多个处理器执行，使得上述一个或多个处理器实现如第一方面中任一实现方式描述的方法。
第四方面，本申请实施例提供了一种计算机可读介质，其上存储有计算机程序，上述程序被处理器执行时实现如第一方面中任一实现方式描述的方法。
本申请实施例提供的图像检测方法和装置，首先对待检测图像进行预设人体部位识别，得到待检测图像中包括的预设人体部位图像的位置信息。从而可以基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像。之后分别对截取的预设人体部位图像和待检测图像进行预设类别图像分类，得到分类结果信息。最后基于分类结果信息，生成用于表征待检测图像是否为预设类别图像的检测结果信息。由于检测结果信息是基于待检测图像和截取的预设人体部位图像生成的。因此，检测结果信息综合了待检测图像的整体信息和局部信息，从而提高了检测结果信息的准确率。
附图说明
通过阅读参照以下附图所作的对非限制性实施例所作的详细描述，本申请的其它特征、目的和优点将会变得更明显：
图1是是本申请的一个实施例可以应用于其中的示例性系统架构图；
图2是根据本申请的图像检测方法的一个实施例的流程图；
图3是根据本申请的图像检测方法的一个应用场景的示意图；
图4是根据本申请的图像检测方法的又一个实施例的流程图；
图5是根据本申请的图像检测装置的一个实施例的结构示意图；
图6是适于用来实现本申请实施例的电子设备的计算机系统的结构示意图。
具体实施方式
下面结合附图和实施例对本申请作进一步的详细说明。可以理解的是，此处所描述的具体实施例仅仅用于解释相关发明，而非对该发明的限定。另外还需要说明的是，为了便于描述，附图中仅示出了与有关发明相关的部分。
需要说明的是，在不冲突的情况下，本申请中的实施例及实施例中的特征可以相互组合。下面将参考附图并结合实施例来详细说明本申请。
图1示出了可以应用本申请实施例的图像检测方法或图像检测装置的示例性系统架构100。
如图1所示，系统架构100可以包括终端101、102、103，网络104和服务器105。网络104用以在终端设备101、102、103和服务器105之间提供通信链路的介质。网络104可以包括各种连接类型，例如有线、无线通信链路或者光纤电缆等等。
终端设备101、102、103通过网络104与服务器105交互，例如将待检测图像发送至服务器105。终端设备101、102、103上可以安装有各类拍照应用、图片处理应用等。
终端设备101、102、103可以是硬件，也可以是软件。当终端设备101、102、103为硬件时，可以是能够拍摄或者存储图像的设备，包括但不限于：照相机、具备拍照功能的手机、图片存储服务器等等。当终端设备101、102、103为软件时，可以安装在上述所列举的电子设备中。其可以实现成多个软件或软件模块(例如用来提供拍照服务)，也可以实现成单个软件或软件模块。在此不做具体限定。
服务器105可以是提供各种服务的服务器，例如对从终端设备101、102、103获取待检测图像进行检测，得到检测结果信息。如果需要，也可以将检测结果信息发送至终端设备101、102、103。
需要说明的是，本申请实施例所提供的图像检测方法可以由服务器105执行，也可以由终端设备执行。相应地，图像检测装置可以设置于服务器105中，也可以设置于终端设备中。
需要说明的是，终端设备101、102、103中也可以对图像进行检测。此时，图像检测方法也可以由终端设备101、102、103执行。相应地，图像检测装置也可以设置于终端设备101、102、103中。此时，示例性系统架构100可以不存在服务器105和网络104。
需要说明的是，服务器可以是硬件，也可以是软件。当服务器为硬件时，可以实现成多个服务器组成的分布式服务器集群，也可以实现成单个服务器。当服务器为软件时，可以实现成多个软件或软件模块(例如用来提供分布式服务)，也可以实现成单个软件或软件模块。在此不做具体限定。
应该理解，图1中的终端设备、网络和服务器的数目仅仅是示意性的。根据实现需要，可以具有任意数目的终端设备、网络和服务器。
继续参考图2，示出了根据本申请的图像检测方法的一个实施例的流程200。该图像检测方法，包括以下步骤：
步骤201，获取待检测图像。
在本实施例中，图像检测方法的执行主体可以通过有线连接方式或者无线连接方式从终端设备获取待检测图像。其中，待检测图像可以是任意图像。待检测图像的确定可以由技术人员指定，也可以根据一定的条件筛选。此外，待检测图像还可以存储于上述执行主体本地。此时，上述执行主体可以直接本地获取待检测图像。
步骤202，对待检测图像进行预设人体部位识别，得到待检测图像中包括的至少一个预设人体部位图像的位置信息。
在本实施例中，上述执行主体可以通过各种方法对待检测图像进行预设人体部位识别，得到待检测图像中包括的至少一个预设人体部位图像的位置信息。其中，预设人体部位可以是人体的至少一个部位，包括但不限于以下至少一项：嘴巴、眼睛、鼻子、女性胸部、女性生殖器官、男性生殖器官等等。其中，位置信息用于表征预设人体部位图像相对于待检测图像的位置。位置信息有各种形式，例如，标注框、坐标等等。
作为示例，上述执行主体可以通过级联分类器进行预设人体部位识别，从而得到待检测图像中包括的至少一个预设人体部位图像的位 置信息。其中，级联分类器可以是级联的多个分类器(例如Haar分类器)。作为示例，OpenCV(Open Source Computer Vision Library，开源的跨平台计算机视觉库)中包括很多针对人体的不同部位(例如，嘴唇、眼睛等)的分类器，可以根据需要选取。多个分类器的检测率不同。检测率可以是检测到图像中显示的预设人体部位的概率。多个分类器可以按照检测率由大到小的顺序进行级联。具体来说，上述执行主体可以将待检测图像输入级联分类器。首先由检测率最大的分类器对待检测图像进行检测，若检测到待检测图像显示预设人体部位，将待检测图像送至下一级分类器。后续的分类器按照同样的方式进行检测，直至检测率最小的一个分类器。若检测到图像中没有显示预设人体部位，可以表征得到图像中没有显示预设人体部位的信息。若最后一级分类器(检测率最小的分类器)检测到图像中显示预设人体部位，可以输出用以表示图像中显示的预设人体部位的位置信息的标注框。从而得到待检测图像中包括的至少一个预设人体部位图像的位置信息。
在本实施例的一些可选的实现方式中，可以将待检测图像输入预先训练的、用于识别预设人体部位图像的预设人体部位识别模型，得到识别结果，识别结果包括待检测图像中包括的至少一个预设人体部位图像的位置信息。其中，预设人体部位识别模型可以用于检测图像中是否包含预设人体部位图像，即图像中是否显示预设人体部位。
在这些实现方式中，预设人体部位识别模型可以通过以下步骤训练得到：
第一步，执行主体可以获取训练样本集合。其中，训练样本集合中的每个训练样本可以包括样本图像和样本图像的标注信息。标注信息包括样本图像中包含的预设人体部位的图像的位置信息和样本图像中显示的预设人体部位的类别信息。其中，位置信息用于表征预设人体部位的图像相对于样本图像的位置。位置信息有各种形式，例如，标注框、坐标等等。预设人体部位的类别信息用于表示预设人体部位的类别。其中，预设人体部位可以是至少一个人体部位。作为示例，预设人体部位可以包括女性胸部、男性生殖器、女性生殖器这三个部 位。那么，预设人体部位的类别信息可以是“00”、“01”、“10”，分别用于表示这三个部位。
第二步，执行主体可以将训练样本集合中的训练样本的样本图像作为输入，将与输入的样本图像对应的标注信息作为期望输出，训练得到预设人体部位检测模型。具体来说，可以将训练样本集合中的训练样本的样本图像输入初始预设人体部位检测模型。其中，初始预设人体部位检测模型可以是各种目标检测网络。作为示例，可以是现有的SSD(Single Shot MultiBox Detector)或YOLO(You Only Look Once)等等。实践中，可以为初始预设人体部位检测模型设置初始值。例如，可以是一些不同的小随机数。“小随机数”用来保证网络不会因权值过大而进入饱和状态，从而导致训练失败，“不同”用来保证网络可以正常地学习。之后，可以得到输入的样本图像的检测结果。以与输入的样本图像的标注信息作为初始预设人体部位检测模型的期望输出，利用机器学习方法训练初始预设人体部位检测模型。具体来说，可以首先利用预设的损失函数计算得到的检测结果与标注信息之间的差异。然后，可以基于所得到的差异，调整初始预设人体部位检测模型的参数，并在满足预设的训练结束条件的情况下，结束训练，并将训练后的初始预设人体部位检测模型作为预设人体部位检测模型。这里的训练结束条件包括但不限于以下至少一项：训练时间超过预设时长；训练次数达到预设次数；计算所得的差异小于预设差异阈值。
这里可以采用各种方式基于所得到的检测结果与输入的训练样本对应的标注信息之间的差异，调整初始预设人体部位检测模型的参数。例如，可以采用BP(Back Propagation，反向传播)算法或者SGD(Stochastic Gradient Descent，随机梯度下降)算法来调整初始图像分类网络的参数。
需要说明的是，训练步骤的执行主体与图像检测方法的执行主体可以相同，也可以不同。若相同，执行主体可以在训练得到预设人体部位检测模型后，将预设人体部位检测模型的网络结构和参数值存储于本地。若不同，训练步骤的执行主体在训练得到预设人体部位检测模型后，可以将模型的网络结构和参数值发送至图像检测方法的执行 主体。
步骤203，基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像。
在本实施例中，上述执行主体基于得到的位置信息，利用各种方法(例如，各种截图应用)从待检测图像中截取预设数量的预设人体部位图像。作为示例，位置信息可以是矩形标注框。实践中，矩形标注框可以用向量表示。向量中可以包含矩形标注框的几何中心的坐标、矩形标注框的高度和宽度。
步骤204，分别对截取的预设人体部位图像和待检测图像进行预设类别图像分类，得到分类结果信息。
在本实施例中，上述执行主体可以利用各种方法分别对截取的预设人体部位图像和待检测图像进行预设类别图像分类，得到分类结果信息。其中，预设类别图像可以是各种类别的图像。例如，可以是面部图像、低俗色情图像等等。需要说明的是，此处的预设类别图像与预设人体部位图像是相匹配的。具体来说，若预设类别图像为面部图像。那么，预设人体部位图像可以是嘴巴图像、眼睛图像等等。若预设类别图像低俗色情图像。那么，预设人体部位图像可以是胸部图像、人体生殖器官图像等等。其中，分类结果信息用于表征图像是否为预设类别图像。分类结果信息可以是各种形式的信息。作为示例，分类结果信息可以是数值。例如，用“0”表示不是预设类别图像，用“1”表示是预设类别图像。作为示例，分类结果信息还可以是文字、字符等等。
作为示例，上述执行主体可以采用分类模型(例如词包模型)，分别对截取的预设人体部位图像和待检测图像进行预设类别图像分类，得到分类结果信息。其中，词包模型应广泛应用于图像识别，其实现可以包括特征提取、特征编码、特征汇聚和利用分类器进行分类。具体来说，特征提取可以利用各种检测算子，例如Harris角点检测算子、FAST(Features from Accelerated Segment Test)算子等，对对象进行特征提取。在此基础上，为了提高特征表达的鲁棒性，可以对特征进行编码。例如，通过查询词典来实现特征编码。之后，对于经过编码的 多个特征进行拼接，并作为图像的最终表达，其具体表现形式可以为向量。最后，利用支持向量机等分类器对拼接得到的向量进行分类。从而得到分类结果信息。
步骤205，基于分类结果信息，生成用于表征待检测图像是否为预设类别图像的检测结果信息。
在本实施例中，上述执行主体可以采用多种方法基于分类结果信息，生成用于表征待检测图像是否为预设类别图像的检测结果信息。
作为示例，步骤204中得到分类结果信息可以包括第一分类结果信息和第二分类结果信息。其中，第一分类结果信息用于表征截取的预设人体部位图像是否为预设类别图像。第二分类结果信息用于表征待检测图像是否为预设类别图像。上述执行主体可以将第一分类结果信息和第二分类结果信息输入分类器。其中，作为示例，分类器可以是Softmax分类器。Softmax可以将输入信息映射到(0,1)区间上。而且在计算过程中会用到所有的输入信息。之后，可以将第二分类结果信息对应的分类器的输出作为检测结果信息。
作为示例，上述执行主体还可以通过查询预设的、分类结果信息与检测结果信息的对应关系表，而得到检测结果信息。其中，对应关系表可以是基于大量的统计得到的。对应关系表征可以记载大量的结果信息以及对应的检测结果信息。上述执行主体可以对于步骤204中得到的分类结果信息，在对应关系表中进行查询，若存在与得到的分类结果信息匹配的分类结果信息，则可以得到匹配的分类结果信息对应的检测结果信息。之后，可以将得到的检测结果信息作为用于表征待检测图像是否为预设类别图像的检测结果信息。
继续参见图3，图3是根据本实施例的图像检测方法的应用场景的一个示意图。在图3的应用场景中，图像检测方法的执行主体可以是服务器300。服务器300首先获取待检测图像301。之后，对待检测图像进行预设人体部位识别。这里的预设人体部位可以是胸部。得到待检测图像301中包括的胸部图像的位置信息，图中显示为标注框302。之后，服务器300从待检测图像301中截取胸部图像，得到截取的图像303。然后分别将截取的图像303和待检测图像301输入分类 模型(例如词包模型)304，得到分类结果信息。如图所示，截取的图像303和待检测图像301的分类结果信息均为“1”。表示截取的图像303和待检测图像301均为预设类别图像(例如不良图像)。在此基础上，服务器300可以通过查询对应关系表305。通过查询，分类结果信息均为“1”对应的检测结果信息也为“1”。从而可以得到用于表征待检测图像是否为预设类别图像(例如不良图像)的检测结果信息。
本申请的上述实施例提供的方法可以首先获取待检测图像。之后，对待检测图像进行预设人体部位识别，得到待检测图像中包括的至少一个预设人体部位图像的位置信息。从而可以基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像。然后分别对截取的预设人体部位图像和待检测图像进行预设类别图像分类，得到分类结果信息。最后，基于分类结果信息，生成用于表征待检测图像是否为预设类别图像的检测结果信息。由于检测结果信息是基于待检测图像和截取的预设人体部位图像生成的。因此，检测结果信息综合了待检测图像的整体信息和局部信息，从而提高了检测结果信息的准确率。
进一步参考图4，其示出了图像检测方法又一个实施例的流程400。该图像检测方法的流程400，包括以下步骤：
步骤401，获取待检测图像。
本实施例中，步骤401的具体实现及所带来的技术效果与图2对应的实施例的步骤201类似，在此不再赘述。
步骤402，将待检测图像输入预先训练的、用于识别预设人体部位图像的预设人体部位识别模型，得到识别结果。
在本实施例中，识别结果包括待检测图像中包括的至少一个预设人体部位图像的位置信息、待检测图像中包括的至少一个预设人体部位图像所显示的人体部位的类别信息和置信度。其中，置信度用于表示类别信息的可信程度。实践中，置信度可以采用概率值表示。
在本实施例中，图像检测方法的执行主体可以将待检测图像输入预先训练的、用于识别预设人体部位图像的预设人体部位识别模型。其中，预设人体部位识别模型用于识别图像中的预设人体部位图像。
步骤403，按照置信度由大到小的顺序，基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像。
在本实施例中，待检测图像中可以包括至少一个预设人体部位图像。因此，上述执行主体可按照置信度由大到小的顺序，基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像。
步骤404，分别将截取的预设人体部位图像和待检测图像输入预先训练的、用于确定图像是否为预设类别图像的预设类别图像分类模型，得到分类结果信息。
在本实施例中，上述执行主体可以分别将截取的预设人体部位图像和待检测图像输入预先训练的预设类别图像分类模型，得到分类结果信息。其中，分类结果信息可以是用于表征图像是否为预设类别图像的信息。需要说明的是，这里的分类结果信息可以包括第一分类结果信息和第二分类结果信息。其中，第一分类结果信息可以用于表征截取的预设人体部位图像是否为预设类别图像。第二分类结果信息可以用于表征待检测图像是否为预设类别图像。
预设类别图像可以是任意类别的图像。作为示例，可以是人脸图像、头部图像、低俗色情图像等等。预设类别图像分类模型用于确定图像是否为预设类别图像。
作为示例，预设类别图像分类模型可以通过以下步骤训练得到：
第一步，获取训练样本集合。其中，训练样本集合中的每个训练样本包括样本图像和标注信息。标注信息用于表征样本图像是否为预设类别图像。这里，标注信息可以是各种形式。作为示例，标注信息可以是数值。例如，用“0”表示不是预设类别图像，用“1”表示是预设类别图像。作为示例，标注信息还可以是文字、字符等等。
第二步，对于训练样本集合中的训练样本，将该训练样本的样本图像作为输入，将与输入的样本图像对应的标注信息作为期望输出，利用机器学习的方法，训练得到预设类别分类模型。具体来说，可以基于训练样本集合对初始预设类别图像分类模型进行训练，得到预设类别图像分类模型。其中，初始预设类别图像分类模型可以是各种图像分类网络。作为示例，可以是残差网络(Deep Residual Network， ResNet)、VGG等等。VGG是某大学的视觉几何小组(Visual Geometry Group，VGG)提出的分类模型。
具体来说，可以将训练样本的样本图像输入初始预设类别分类模型。实践中，可以为初始预设类别分类模型设置初始值。之后，可以得到输入的样本图像的分类结果信息。然后可以利用预设的损失函数计算得到的分类结果信息与标注信息之间的差异。之后，可以基于得到的差异，调整初始预设类别分类模型的参数，并在满足预设的训练结束条件的情况下，结束训练，将训练后的初始预设类别分类模型作为预设类别图像分类模型。这里的训练结束条件包括但不限于以下至少一项：训练时间超过预设时长；训练次数达到预设次数；计算所得的差异小于预设差异阈值。
步骤405，基于分类结果信息，生成用于表征待检测图像是否为预设类别图像的检测结果信息。
在本实施例中，步骤405的具体实现及所带来的技术效果与图2对应的实施例中的步骤205类似，在此不再赘述。
从图4中可以看出，与图2对应的实施例相比，本实施例中的图像检测方法通过将待检测图像输入预设人体部位识别模型，得到识别结果。并按照置信度由大到小的顺序，截取预设数量的预设人体部位图像。与其他识别方法相比，提高了预设人体部位识别的准确性。由此，提高了待检测图像的检测准确率。
进一步参考图5，作为对上述各图所示方法的实现，本申请提供了一种图像检测装置的一个实施例，该装置实施例与图2所示的方法实施例相对应，该装置具体可以应用于各种电子设备中。
如图5所示，本实施例的图像检测装置500包括：获取单元501、识别单元502、截取单元503、分类单元504和生成单元505。其中，获取单元501被配置成获取待检测图像。识别单元502被配置成对待检测图像进行预设人体部位识别，得到待检测图像中包括的至少一个预设人体部位图像的位置信息。截取单元503被配置成基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像。分类单元504被配置成分别对截取的预设人体部位图像和待检测图像进行预 设类别图像分类，得到分类结果信息。生成单元505被配置成基于分类结果信息，生成用于表征待检测图像是否为预设类别图像的检测结果信息。
本实施例中的图像检测装置600包括的获取单元501、识别单元502、截取单元503、分类单元504和生成单元505的具体实现及所带来的技术效果与图2对应的实施例的步骤201-205，在此不再赘述。
在本实施例的一些可选的实现方式中，识别单元进一步被配置成：将待检测图像输入预先训练的、用于识别预设人体部位图像的预设人体部位识别模型，得到识别结果，识别结果包括待检测图像中包括的至少一个预设人体部位图像的位置信息。
在本实施例的一些可选的实现方式中，识别结果还可以包括：待检测图像中包括的至少一个预设人体部位图像所显示的人体部位的类别信息和置信度。
在本实施例的一些可选的实现方式中，截取单元503可以进一步被配置成：对于至少一个预设人体部位图像，按照置信度由大到小的顺序，基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像。
在本实施例的一些可选的实现方式中，分类单元504可以进一步被配置成：分别将截取的预设人体部位图像和待检测图像输入预先训练的、用于确定图像是否为预设类别图像的预设类别图像分类模型，得到分类结果信息。
在本实施例中，上述图像检测装置可以首先通过获取单元501获取待检测图像。之后，识别单元502对待检测图像进行预设人体部位识别，得到待检测图像中包括的至少一个预设人体部位图像的位置信息。从而截取单元503可以基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像。然后分类单元504分别对截取的预设人体部位图像和待检测图像进行预设类别图像分类，得到分类结果信息。最后，生成单元505基于分类结果信息，生成用于表征待检测图像是否为预设类别图像的检测结果信息。由于检测结果信息是基于待检测图像和截取的预设人体部位图像生成的。因此，检测结果信息 综合了待检测图像的整体信息和局部信息，从而提高了检测结果信息的准确率。
继续参考图6，其示出了适于用来实现本申请实施例的电子设备的计算机系统600的结构示意图。图6示出的电子设备仅仅是一个示例，不应对本申请实施例的功能和使用范围带来任何限制。
如图6所示，计算机系统600包括中央处理单元(CPU)601，其可以根据存储在只读存储器(ROM)602中的程序或者从存储部分608加载到随机访问存储器(RAM)603中的程序而执行各种适当的动作和处理。在RAM 603中，还存储有系统600操作所需的各种程序和数据。CPU 601、ROM 602以及RAM 603通过总线604彼此相连。输入/输出(I/O)接口605也连接至总线604。
以下部件连接至I/O接口605：包括键盘、鼠标等的输入部分606；包括诸如阴极射线管(CRT)、液晶显示器(LCD)等以及扬声器等的输出部分607；包括硬盘等的存储部分608；以及包括诸如LAN卡、调制解调器等的网络接口卡的通信部分609。通信部分609经由诸如因特网的网络执行通信处理。驱动器610也根据需要连接至I/O接口605。可拆卸介质611，诸如磁盘、光盘、磁光盘、半导体存储器等等，根据需要安装在驱动器610上，以便于从其上读出的计算机程序根据需要被安装入存储部分608。
特别地，根据本公开的实施例，上文参考流程图描述的过程可以被实现为计算机软件程序。例如，本公开的实施例包括一种计算机程序产品，其包括承载在计算机可读介质上的计算机程序，该计算机程序包含用于执行流程图所示的方法的程序代码。在这样的实施例中，该计算机程序可以通过通信部分609从网络上被下载和安装，和/或从可拆卸介质611被安装。在该计算机程序被中央处理单元(CPU)601执行时，执行本申请的方法中限定的上述功能。
需要说明的是，本申请所述的计算机可读介质可以是计算机可读信号介质或者计算机可读存储介质或者是上述两者的任意组合。计算机可读存储介质例如可以是——但不限于——电、磁、光、电磁、红 外线、或半导体的系统、装置或器件，或者任意以上的组合。计算机可读存储介质的更具体的例子可以包括但不限于：具有一个或多个导线的电连接、便携式计算机磁盘、硬盘、随机访问存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、光纤、便携式紧凑磁盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。在本申请中，计算机可读存储介质可以是任何包含或存储程序的有形介质，该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。而在本申请中，计算机可读的信号介质可以包括在基带中或者作为载波一部分传播的数据信号，其中承载了计算机可读的程序代码。这种传播的数据信号可以采用多种形式，包括但不限于电磁信号、光信号或上述的任意合适的组合。计算机可读的信号介质还可以是计算机可读存储介质以外的任何计算机可读介质，该计算机可读介质可以发送、传播或者传输用于由指令执行系统、装置或者器件使用或者与其结合使用的程序。计算机可读介质上包含的程序代码可以用任何适当的介质传输，包括但不限于：无线、电线、光缆、RF等等，或者上述的任意合适的组合。
可以以一种或多种程序设计语言或其组合来编写用于执行本申请的操作的计算机程序代码，所述程序设计语言包括面向对象的程序设计语言—诸如Java、Smalltalk、C++，还包括常规的过程式程序设计语言—诸如“C”语言或类似的程序设计语言。程序代码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)。
附图中的流程图和框图，图示了按照本申请各种实施例的系统、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、或代码的一部分，该模块、程序段、或代码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。也应当注意，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，两个接连地表示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操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。
描述于本申请实施例中所涉及到的单元可以通过软件的方式实现，也可以通过硬件的方式来实现。所描述的单元也可以设置在处理器中，例如，可以描述为：一种处理器包括获取单元、识别单元、截取单元、分类单元和生成单元。其中，这些单元的名称在某种情况下并不构成对该单元本身的限定，例如，获取单元还可以被描述为“获取待检测图像的单元”。
作为另一方面，本申请还提供了一种计算机可读介质，该计算机可读介质可以是上述实施例中描述的电子设备中所包含的；也可以是单独存在，而未装配入该电子设备中。上述计算机可读介质承载有一个或者多个程序，当上述一个或者多个程序被该电子设备执行时，使得该电子设备：获取待检测图像；对待检测图像进行预设人体部位识别，得到待检测图像中包括的至少一个预设人体部位图像的位置信息；基于得到的位置信息，从待检测图像中截取预设数量的预设人体部位图像；分别对截取的预设人体部位图像和待检测图像进行预设类别图像分类，得到分类结果信息；基于分类结果信息，生成用于表征待检测图像是否为预设类别图像的检测结果信息。
以上描述仅为本申请的较佳实施例以及对所运用技术原理的说明。本领域技术人员应当理解，本申请中所涉及的发明范围，并不限于上述技术特征的特定组合而成的技术方案，同时也应涵盖在不脱离上述发明构思的情况下，由上述技术特征或其等同特征进行任意组合而形成的其它技术方案。例如上述特征与本申请中公开的(但不限于)具有类似功能的技术特征进行互相替换而形成的技术方案。

Claims (12)

  1. 一种图像检测方法，包括：
    获取待检测图像；
    对所述待检测图像进行预设人体部位识别，得到所述待检测图像中包括的至少一个预设人体部位图像的位置信息；
    基于得到的位置信息，从所述待检测图像中截取预设数量的预设人体部位图像；
    分别对截取的预设人体部位图像和所述待检测图像进行预设类别图像分类，得到分类结果信息；
    基于所述分类结果信息，生成用于表征所述待检测图像是否为预设类别图像的检测结果信息。
  2. 根据权利要求1所述的方法，其中，所述对所述待检测图像进行预设人体部位识别，得到所述待检测图像中包括的至少一个预设人体部位图像的位置信息，包括：
    将所述待检测图像输入预先训练的、用于识别预设人体部位图像的预设人体部位识别模型，得到识别结果，所述识别结果包括所述待检测图像中包括的至少一个预设人体部位图像的位置信息。
  3. 根据权利要求2所述的方法，其中，所述识别结果还包括：
    所述待检测图像中包括的至少一个预设人体部位图像所显示的人体部位的类别信息和置信度。
  4. 根据权利要求3所述的方法，其中，所述基于得到的位置信息，从所述待检测图像截取预设数量的预设人体部位图像，包括：
    按照置信度由大到小的顺序，基于得到的位置信息，从所述待检测图像中截取预设数量的预设人体部位图像。
  5. 根据权利要求1-4中任一所述的方法，其中，所述分别对截取 的预设人体部位图像和所述待检测图像进行预设类别图像分类，得到分类结果信息，包括：
    分别将截取的预设人体部位图像和所述待检测图像输入预先训练的、用于确定图像是否为预设类别图像的预设类别图像分类模型，得到分类结果信息。
  6. 一种图像检测装置，包括：
    获取单元，被配置成获取待检测图像；
    识别单元，被配置成对所述待检测图像进行预设人体部位识别，得到所述待检测图像中包括的至少一个预设人体部位图像的位置信息；
    截取单元，被配置成基于得到的位置信息，从所述待检测图像中截取预设数量的预设人体部位图像；
    分类单元，被配置成分别对截取的预设人体部位图像和所述待检测图像进行预设类别图像分类，得到分类结果信息；
    生成单元，被配置成基于所述分类结果信息，生成用于表征所述待检测图像是否为预设类别图像的检测结果信息。
  7. 根据权利要求6所述的装置，其中，所述识别单元进一步被配置成：
    将所述待检测图像输入预先训练的、用于识别预设人体部位图像的预设人体部位识别模型，得到识别结果，所述识别结果包括所述待检测图像中包括的至少一个预设人体部位图像的位置信息。
  8. 根据权利要求7所述的装置，其中，所述识别结果还包括：
    所述待检测图像中包括的至少一个预设人体部位图像所显示的人体部位的类别信息和置信度。
  9. 根据权利要求8所述的装置，其中，所述截取单元进一步被配置成：
    对于所述至少一个预设人体部位图像，按照置信度由大到小的顺序，基于得到的位置信息，从所述待检测图像中截取预设数量的预设人体部位图像。
  10. 根据权利要求6-9中任一所述的装置，其中，所述分类单元进一步被配置成：
    分别将截取的预设人体部位图像和所述待检测图像输入预先训练的、用于确定图像是否为预设类别图像的预设类别图像分类模型，得到分类结果信息。
  11. 一种电子设备，包括：
    一个或多个处理器；
    存储装置，其上存储有一个或多个程序，
    当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现如权利要求1-5中任一所述的方法。
  12. 一种计算机可读介质，其上存储有计算机程序，其中，所述程序被处理器执行时实现如权利要求1-5中任一所述的方法。
PCT/CN2018/116338 2018-07-06 2018-11-20 图像检测方法和装置 WO2020006964A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810734680.8A CN109558779A (zh) 2018-07-06 2018-07-06 图像检测方法和装置
CN201810734680.8 2018-07-06

Publications (1)

Publication Number Publication Date
WO2020006964A1 true WO2020006964A1 (zh) 2020-01-09

Family

ID=65864446

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/116338 WO2020006964A1 (zh) 2018-07-06 2018-11-20 图像检测方法和装置

Country Status (2)

Country Link
CN (1) CN109558779A (zh)
WO (1) WO2020006964A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111291640A (zh) * 2020-01-20 2020-06-16 北京百度网讯科技有限公司 用于识别步态的方法和装置
CN111539456A (zh) * 2020-04-02 2020-08-14 浙江华睿科技有限公司 一种目标识别方法及设备
CN111935106A (zh) * 2020-07-23 2020-11-13 湖北工业大学 一种利用改进验证码收集带标签的数据集的方法
CN112348615A (zh) * 2020-03-24 2021-02-09 北京沃东天骏信息技术有限公司 用于审核信息的方法和装置
CN112580745A (zh) * 2020-12-29 2021-03-30 北京五八信息技术有限公司 图像识别方法、装置、电子设备和计算机可读介质

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110310142A (zh) * 2019-05-20 2019-10-08 浙江新再灵科技股份有限公司 基于人群价值分析的电梯广告投放方法及装置
CN111814852A (zh) * 2020-06-24 2020-10-23 理光软件研究所（北京）有限公司 图像检测方法、装置、电子设备和计算机可读存储介质
KR20220169373A (ko) * 2021-06-17 2022-12-27 센스타임 인터내셔널 피티이. 리미티드. 타겟 검출 방법들, 장치들, 전자 디바이스들 및 컴퓨터 판독가능한 저장 매체
WO2022263904A1 (en) * 2021-06-17 2022-12-22 Sensetime International Pte. Ltd. Target detection methods, apparatuses, electronic devices and computer-readable storage media

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102184419A (zh) * 2011-04-13 2011-09-14 深圳市迈科龙影像技术有限公司 基于敏感部位检测的色情图像识别方法
CN103366160A (zh) * 2013-06-28 2013-10-23 西安交通大学 融合肤色、人脸和敏感部位检测的不良图像判别方法
EP2657857A1 (en) * 2012-04-27 2013-10-30 ATG Advanced Swiss Technology Group AG Method for binary classification of a query image
CN105844283A (zh) * 2015-01-16 2016-08-10 阿里巴巴集团控股有限公司 用于识别图像类目归属的方法、图像搜索方法及装置
CN107992783A (zh) * 2016-10-26 2018-05-04 上海银晨智能识别科技有限公司 人脸图像处理方法及装置

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN100481112C (zh) * 2007-06-11 2009-04-22 湖北东润科技有限公司 一种人脸识别系统与识别方法
CN101763502B (zh) * 2008-12-24 2012-07-25 中国科学院自动化研究所 一种高效的敏感图像检测方法及其系统
CN103049733B (zh) * 2011-10-11 2016-02-03 株式会社理光 人脸检测方法和人脸检测设备
CN107871134A (zh) * 2016-09-23 2018-04-03 北京眼神科技有限公司 一种人脸检测方法及装置
CN108229575A (zh) * 2018-01-19 2018-06-29 百度在线网络技术（北京）有限公司 用于检测目标的方法和装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102184419A (zh) * 2011-04-13 2011-09-14 深圳市迈科龙影像技术有限公司 基于敏感部位检测的色情图像识别方法
EP2657857A1 (en) * 2012-04-27 2013-10-30 ATG Advanced Swiss Technology Group AG Method for binary classification of a query image
CN103366160A (zh) * 2013-06-28 2013-10-23 西安交通大学 融合肤色、人脸和敏感部位检测的不良图像判别方法
CN105844283A (zh) * 2015-01-16 2016-08-10 阿里巴巴集团控股有限公司 用于识别图像类目归属的方法、图像搜索方法及装置
CN107992783A (zh) * 2016-10-26 2018-05-04 上海银晨智能识别科技有限公司 人脸图像处理方法及装置

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111291640A (zh) * 2020-01-20 2020-06-16 北京百度网讯科技有限公司 用于识别步态的方法和装置
CN111291640B (zh) * 2020-01-20 2023-02-17 北京百度网讯科技有限公司 用于识别步态的方法和装置
CN112348615A (zh) * 2020-03-24 2021-02-09 北京沃东天骏信息技术有限公司 用于审核信息的方法和装置
CN111539456A (zh) * 2020-04-02 2020-08-14 浙江华睿科技有限公司 一种目标识别方法及设备
CN111539456B (zh) * 2020-04-02 2024-03-01 浙江华睿科技股份有限公司 一种目标识别方法及设备
CN111935106A (zh) * 2020-07-23 2020-11-13 湖北工业大学 一种利用改进验证码收集带标签的数据集的方法
CN111935106B (zh) * 2020-07-23 2022-05-13 湖北工业大学 一种利用改进验证码收集带标签的数据集的方法
CN112580745A (zh) * 2020-12-29 2021-03-30 北京五八信息技术有限公司 图像识别方法、装置、电子设备和计算机可读介质

Also Published As

Publication number Publication date
CN109558779A (zh) 2019-04-02

Similar Documents

Publication Publication Date Title
WO2020006964A1 (zh) 图像检测方法和装置
WO2020006963A1 (zh) 生成图像检测模型的方法和装置
CN109726624B (zh) 身份认证方法、终端设备和计算机可读存储介质
WO2020000879A1 (zh) 图像识别方法和装置
US11270099B2 (en) Method and apparatus for generating facial feature
JP6030240B2 (ja) 顔認識のための方法および装置
US11436863B2 (en) Method and apparatus for outputting data
CN111368685B (zh) 关键点的识别方法、装置、可读介质和电子设备
WO2020019591A1 (zh) 用于生成信息的方法和装置
EP3872652A2 (en) Method and apparatus for processing video, electronic device, medium and product
CN108494778A (zh) 身份认证方法和装置
WO2022105118A1 (zh) 基于图像的健康状态识别方法、装置、设备及存储介质
CN109670444B (zh) 姿态检测模型的生成、姿态检测方法、装置、设备及介质
US11734954B2 (en) Face recognition method, device and electronic equipment, and computer non-volatile readable storage medium
CN108229375B (zh) 用于检测人脸图像的方法和装置
US20230036338A1 (en) Method and apparatus for generating image restoration model, medium and program product
CN108388889B (zh) 用于分析人脸图像的方法和装置
WO2022127480A1 (zh) 一种人脸识别方法和相关设备
CN110059624A (zh) 用于检测活体的方法和装置
CN115376559A (zh) 基于音视频的情绪识别方法、装置及设备
CN111881740A (zh) 人脸识别方法、装置、电子设备及介质
CN111931628A (zh) 人脸识别模型的训练方法、装置及相关设备
CN111783674A (zh) 一种基于ar眼镜的人脸识别方法和系统
US20210295016A1 (en) Living body recognition detection method, medium and electronic device
WO2024077971A1 (zh) 活体检测方法及装置

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: 18925301

Country of ref document: EP

Kind code of ref document: A1

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 14.05.2021)

122 Ep: pct application non-entry in european phase

Ref document number: 18925301

Country of ref document: EP

Kind code of ref document: A1