CN114067406A - Key point detection method, device, equipment and readable storage medium - Google Patents
Key point detection method, device, equipment and readable storage medium Download PDFInfo
- Publication number
- CN114067406A CN114067406A CN202111371392.9A CN202111371392A CN114067406A CN 114067406 A CN114067406 A CN 114067406A CN 202111371392 A CN202111371392 A CN 202111371392A CN 114067406 A CN114067406 A CN 114067406A
- Authority
- CN
- China
- Prior art keywords
- key point
- target
- key
- image
- coordinates
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The application discloses a key point detection method, a key point detection device and a readable storage medium, wherein the acquired image to be detected is processed to obtain thermodynamic diagrams with preset quantity, each pixel point in the thermodynamic diagrams is distributed with a score, for each thermodynamic diagram, a key point is determined according to the score distributed by each pixel point, corresponding key point information is obtained, a target key point is obtained through screening based on each key point information, coordinates of the target key point are corrected, and the corrected target key point coordinates are obtained. When the key point detection is carried out, when the image only contains one part of information, the other part of information does not exist in the image, so that the key points predicted by the other part of information should not exist in the image in the preset number of key points, the key points which should not exist in the image can be screened out by using the key point information, and the coordinates of the rest key points are corrected, so that the corrected target key point coordinates are obtained.
Description
Technical Field
The present application relates to the field of image recognition and processing, and in particular, to a method, an apparatus, a device, and a readable storage medium for detecting a keypoint.
Background
With the continuous development of science and technology, the image recognition technology is also widely applied, and information support can be provided for various applications by acquiring key features of target objects in images. For example, in the field of face recognition, if a face needs to be recognized more accurately, several special positions on the face need to be detected, so as to obtain information that can distinguish different faces; in the field of automatic driving, if the motion state of a pedestrian needs to be detected, joints of a human body need to be detected so as to analyze the current posture of the human body. Therefore, if more information of the target object in the image needs to be acquired, the key points of the target object need to be further detected, and therefore, how to detect the key points in the image is a problem which is always concerned.
Disclosure of Invention
In view of the above, the present application provides a method, an apparatus, a device and a readable storage medium for detecting keypoints in an image.
In order to achieve the above object, the following solutions are proposed:
a keypoint detection method comprising:
acquiring an image to be detected;
processing an image to be detected to obtain thermodynamic diagrams with preset quantity, wherein each pixel point in the thermodynamic diagrams is distributed with a score, and the score represents the probability that each pixel point is a key point;
determining a key point according to the value distributed by each pixel point aiming at each thermodynamic diagram, and obtaining corresponding key point information;
and screening to obtain target key points based on the information of each key point, and correcting the coordinates of the target key points to obtain corrected coordinates of the target key points.
Optionally, the determining a key point according to the score assigned to each pixel point includes:
and taking the pixel point with the maximum score in each thermodynamic diagram as a key point.
Optionally, the step of processing the image to be detected until the coordinates of the target key points are obtained by screening is completed by using a pre-trained target detection model;
the target detection model is obtained by training with a training image marked with target key point coordinates as training data.
Optionally, the target detection model includes: the method comprises the following steps that an input layer, a thermodynamic diagram acquisition layer, a key point information determination layer and a key point screening layer are sequentially cascaded;
the training process of the target detection model comprises the following steps:
acquiring a training image through an input layer;
processing the training images through the thermodynamic diagram acquisition layer to obtain thermodynamic diagrams with preset quantity;
determining a key point according to the value distributed by each pixel point and aiming at each thermodynamic diagram through a key point information determination layer, and obtaining corresponding key point information;
screening to obtain target key points through a key point screening layer based on the information of each key point, and correcting the coordinates of the target key points to obtain corrected coordinates of the target key points;
and updating parameters of a target detection model by taking the corrected target key point coordinates approaching to the corresponding target key point coordinate labels in the training image as training targets.
Optionally, the key point information includes: keypoint coordinates and keypoint scores.
A keypoint detection device comprising:
the image acquisition unit is used for acquiring an image to be detected;
the thermodynamic diagram determining unit is used for processing the image to be detected to obtain thermodynamic diagrams with preset quantity, wherein each pixel point in the thermodynamic diagrams is distributed with a score, and the score represents the probability that each pixel point is a key point;
the key point information determining unit is used for determining a key point according to the value distributed by each pixel point aiming at each thermodynamic diagram and obtaining corresponding key point information;
and the target key point determining unit is used for screening to obtain target key points based on the information of each key point, and correcting the coordinates of the target key points to obtain corrected coordinates of the target key points.
Optionally, the key point information determining unit executes a process of determining a key point according to the score assigned to each pixel point, including:
and taking the pixel point with the maximum score in each thermodynamic diagram as a key point.
Optionally, the steps executed by the thermodynamic diagram determining unit, the key point information determining unit, and the target key point determining unit are completed by using a pre-trained target detection model, where the target detection model is obtained by training using a training image labeled with target key point coordinates as training data, and the key point detecting device further includes:
the target detection model training unit is used for training to obtain the target detection model, wherein the target detection model comprises:
the method comprises the following steps that an input layer, a thermodynamic diagram acquisition layer, a key point information determination layer and a key point screening layer are sequentially cascaded;
the training process of the target detection model comprises the following steps:
acquiring a training image through an input layer;
processing the training images through the thermodynamic diagram acquisition layer to obtain thermodynamic diagrams with preset quantity;
determining a key point according to the value distributed by each pixel point and aiming at each thermodynamic diagram through a key point information determination layer, and obtaining corresponding key point information;
screening to obtain target key points through a key point screening layer based on the information of each key point, and correcting the coordinates of the target key points to obtain corrected coordinates of the target key points;
and updating parameters of a target detection model by taking the corrected target key point coordinates approaching to the corresponding target key point coordinate labels in the training image as training targets.
A keypoint detection apparatus comprising: a memory and a processor;
the memory is used for storing programs;
the processor is configured to execute the program to implement the steps of the aforementioned keypoint detection method.
A readable storage medium, having stored thereon a computer program which, when being executed by a processor, carries out the steps of the keypoint detection method as described above.
It can be seen from the foregoing technical solutions that, in the method, the apparatus, the device, and the readable storage medium for detecting a key point provided in the embodiments of the present application, an acquired image to be detected is processed to obtain a preset number of thermodynamic diagrams, each pixel point in the thermodynamic diagrams is assigned with a score representing a probability that each pixel point is a key point, for each thermodynamic diagram, a key point is determined according to the assigned score of each pixel point, and corresponding key point information is obtained, a target key point is obtained by screening based on each key point information, and a coordinate of the target key point is corrected to obtain a corrected target key point coordinate.
Further, when the key point detection is performed, the image to be detected may be processed to obtain a preset number of thermodynamic diagrams, and a key point is determined for each thermodynamic diagram to obtain a preset number of key points, but when the image to be detected only includes a part of information, for example, only half of the face is in the image to be detected, since another part of the information does not actually exist in the image to be detected, the key point predicted for another part of the information should not exist in the image to be detected, in the obtained preset number of key points, the key point information corresponding to the key point is obtained, and the key point information that should not exist in the image to be detected may be screened out by using the key point information, and the coordinate of the target key point is corrected, so as to obtain the corrected target key point coordinate.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a flowchart of a method for detecting a key point according to an embodiment of the present disclosure;
FIG. 2 is a flowchart of an alternative method for training a target detection model according to an embodiment of the present disclosure;
fig. 3 is a schematic structural diagram of a key point detection device according to an embodiment of the present disclosure;
fig. 4 is a block diagram of a hardware structure of a keypoint detection apparatus disclosed in an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Fig. 1 is a flowchart of a method for detecting a keypoint according to an embodiment of the present disclosure, where the method includes the following steps:
and S100, acquiring an image to be detected.
Specifically, the image to be detected may be a picture, or may be a frame of image in a video. The mode of obtaining the image to be detected can receive the image directly transmitted by the image acquisition device, such as an image returned by a camera in real time, and can also be a video or an image which is read from other storage equipment and is stored in advance.
And S101, processing the image to be detected to obtain thermodynamic diagrams with preset quantity.
And the scores represent the probability that each pixel point is a key point.
Specifically, after the image to be detected is obtained, a target object to be subjected to the key point detection in the image to be detected can be determined, and the number of preset key points to be detected of the target object can be determined according to the target object, wherein each key point represents a key point category. For example, if the image to be detected includes key points of a human face, the number of the preset key points to be detected of the human face may be determined to be 68, that is, 68 key point categories.
And processing the image to be detected by combining the information obtained from the image to be detected after obtaining the image to be detected, so as to obtain thermodynamic diagrams with preset quantity. In the processing process of the image to be detected, a thermodynamic diagram is obtained for each key point, and each pixel point in the thermodynamic diagram is assigned with a score for representing the probability that each pixel point is a key point, so that the number of the thermodynamic diagrams finally obtained is consistent with the number of the key points to be detected in the image to be detected.
And S102, determining a key point according to the value distributed by each pixel point aiming at each thermodynamic diagram, and obtaining corresponding key point information.
Specifically, after the preset number of thermodynamic diagrams are obtained through the above steps, a key point can be determined for each thermodynamic diagram according to the value assigned to each pixel point. The method for determining the key point may be to select a pixel point with the largest score in each thermodynamic diagram as the key point.
After the key points are determined in the above manner, corresponding key point information can be obtained, the number of the obtained key points is consistent with the number of the thermodynamic diagrams, and each thermodynamic diagram corresponds to one key point and key point information.
And S103, screening to obtain target key points based on the information of the key points, and correcting the coordinates of the target key points to obtain corrected coordinates of the target key points.
Specifically, when the key point detection is performed, in the obtained image to be detected, the target object to be subjected to the key point detection may not be a complete object but only a part of the target object, for example, the target object is a face, but only a left half of the face exists in the image to be detected. In this case, since the number of the key points is predetermined according to the target object, the number of the key points to be predicted cannot be adjusted because the image to be detected only includes a part of the target object, so that a thermodynamic diagram with a preset number and corresponding key points are finally obtained, which causes the key points that should be located in another part of the target object to also appear in the image to be detected, so that the key points belong to wrong key points, and therefore, the target key points need to be screened according to the information of each key point, and the target key points are the key points existing in the part of the target object included in the image to be detected. In this case, the obtained target key points may have a position deviation, for example, the target key points are squeezed together, so that the coordinates of the target key points need to be corrected based on the information of the key points, and finally, the corrected coordinates of the target key points are obtained.
For example, the number of the key points of the face set first is 4, and the key points to be detected are distributed on the left half face and the right half face of the 4 key points, which are the left eye, the right eye, the left mandible and the right mandible respectively. Assuming that the acquired image to be detected contains a left half human face, 4 thermodynamic diagrams and key points and key point information corresponding to each thermodynamic diagram can still be obtained after the image to be detected is processed, but only the left half human face is contained in the image to be detected, so when the right eye and the right lower jaw of the key points distributed on the right half human face are detected, only one point is found on the left half human face contained in the image to be detected as a key point, the obtained key point is not the key point on the right half human face, so that a part of key points can exist and can not appear on the image to be detected, at this time, based on each key point information, the key points which should not appear on the image to be detected can be deleted, only the predicted key points of the left eye and the left lower jaw are reserved, and the coordinates of the key points of the left eye and the left lower jaw are corrected by using the key point information of the left eye and the left lower jaw, and obtaining the coordinates of the key points of the left eye and the left mandible after correction.
According to the technical scheme, the key point detection method provided by the embodiment of the application processes the acquired image to be detected to obtain the thermodynamic diagrams with the preset number, each pixel point in the thermodynamic diagrams is allocated with the value representing the probability that each pixel point is the key point, for each thermodynamic diagram, one key point is determined according to the value allocated by each pixel point, corresponding key point information is obtained, the target key point is obtained through screening based on each key point information, and the coordinate of the target key point is corrected to obtain the corrected coordinate of the target key point.
Further, when the key point detection is performed, the image to be detected may be processed to obtain a preset number of thermodynamic diagrams, and a key point is determined for each thermodynamic diagram to obtain a preset number of key points, but when the image to be detected only includes a part of information, for example, only half of the face is in the image to be detected, since another part of the information does not actually exist in the image to be detected, the key point predicted for another part of the information should not exist in the image to be detected, in the obtained preset number of key points, the key point information corresponding to the key point is obtained, and the key point information that should not exist in the image to be detected may be screened out by using the key point information, and the coordinate of the target key point is corrected, so as to obtain the corrected target key point coordinate.
Further, in some embodiments of the present application, the key point information may include: keypoint coordinates and keypoint scores. The key point score can be the score assigned to the key point in the thermodynamic diagram corresponding to the key point.
At present, the key point detection belongs to one of important research directions in the field of computer vision, a neural network model can be adopted to detect the key points, but when a common key point detection model is used for detecting the key points of target objects which turn over randomly, the detection accuracy can be greatly reduced, so that a target detection model framework can be adopted, each key point is considered as different targets to be detected, and the detection accuracy of the key point detection under different angles is improved. Based on this, in some embodiments of the present application, the steps performed in steps S101-S103 described above may be performed by using a pre-trained target detection model.
Specifically, the target detection model is obtained by training using a training image labeled with target key point coordinates as training data, and can process an image to be detected input into the target detection model and finally output the target key point coordinates. The method comprises the steps of processing an image to be detected in a target detection model to obtain thermodynamic diagrams with preset quantity, determining a key point according to a value distributed by each pixel point for each thermodynamic diagram, obtaining corresponding key point information, screening to obtain a target key point based on each key point information, and correcting coordinates of the target key point to obtain corrected coordinates of the target key point.
The target detection models can be distinguished according to different detection objects, for example, for face key point detection, human body key point detection and vehicle key point detection, different target detection models can be used for detection, the number and types of key points to be detected can be predetermined according to the detection objects, for example, for face key point detection, the number and types of the key points to be detected can be predetermined as 68, and the types of the 68 key points can be predetermined, for example, the types of the key points are key points below the left eyebrow center, the right eyebrow center, the left eye center, the right eye center, the left nose corner, the right nose corner, the left mouth corner, the right mouth corner, the left earlobe, the right earlobe below, the lower chin lowest point and the like.
In the above embodiment, the target detection model is used to detect the key points in the image to be detected, each key point can be regarded as a different target to be detected, and the coordinates of the target key points output by the target detection model are finally obtained, so that the detection of the key points is realized.
In some embodiments of the present application, an alternative architecture for the target detection model may include: the system comprises an input layer, a thermodynamic diagram acquisition layer, a key point information determination layer and a key point screening layer which are sequentially cascaded.
The thermodynamic diagram acquisition layer can output thermodynamic diagrams with preset quantity, the key point information determination layer can output key points and key point information corresponding to each thermodynamic diagram, and the key point screening layer can output target key point coordinates.
Referring to fig. 2, a description is given of an alternative training process of a target detection model provided in an embodiment of the present application, where the training process of the target detection model may include:
and step S200, acquiring a training image through an input layer.
Specifically, after the training image is acquired by the input layer, the training image may be vectorized, and the vectorized training image is output by the thermodynamic diagram acquisition layer, where coordinates of the target key point are labeled in advance in the training image.
Step S201, the training images are processed through the thermodynamic diagram acquisition layer to obtain thermodynamic diagrams in preset quantity.
Specifically, when the thermodynamic diagram acquisition layer processes the training image, the probability that each pixel point is a key point can be analyzed for each pixel point, so that corresponding scores are distributed to the pixel points, corresponding scores can be distributed to the pixel points in a Gaussian distribution mode, and finally obtained scores can be regarded as the probability that the pixel points are the key points.
Step S202, determining a key point according to the value distributed by each pixel point and aiming at each thermodynamic diagram through a key point information determination layer, and obtaining corresponding key point information.
Specifically, a preset number of thermodynamic diagrams can be obtained through the thermodynamic diagram obtaining layer, scores are distributed to all pixel points in the thermodynamic diagrams and used for representing the probability that the pixel points are key points, one key point can be determined from each thermodynamic diagram according to the scores distributed to all the pixel points, and corresponding key point information is obtained.
And S203, screening to obtain target key points through the key point screening layer based on the information of each key point, and correcting the coordinates of the target key points to obtain corrected coordinates of the target key points.
Specifically, when performing the keypoint detection, in the obtained training image, the target object to be subjected to the keypoint detection may not be a complete object but only a part of the target object, for example, the target object is a face, but only a left half of the face exists in the training image. In this case, since the number of the key points is predetermined according to the target object, the number of the key points to be predicted is not adjusted because the training image only includes a part of the target object, so that a predetermined number of thermodynamic diagrams and corresponding key points are finally obtained, which may cause the key points that should be located in another part of the target object to also appear in the training image, and therefore the key points of this kind belong to wrong key points, and therefore the target key points need to be obtained by screening according to the information of each key point, and the target key points are the key points existing in the part of the target object included in the training image. In this case, the obtained target key points may have a position deviation, for example, the target key points are squeezed together, so that the coordinates of the target key points need to be corrected based on the information of the key points, and finally, the corrected coordinates of the target key points are obtained.
And step S204, updating parameters of the target detection model by taking the corrected target key point coordinate approaching to the corresponding target key point coordinate label in the training image as a training target.
In the training process of the target detection model provided in the above embodiment, the corrected coordinates of the target key points approach to the corresponding target key point coordinate labels in the training image as the training targets, the target detection model is trained, and when the determined coordinates of the target key points are inconsistent with the corresponding coordinates of the key points in the training image, the parameters of the target detection model are updated, so that the model can select the target key points more accurately.
The following describes the key point detection device provided in the embodiment of the present application, and the key point detection device described below and the key point detection method described above may be referred to in correspondence with each other.
Fig. 3 is a schematic structural diagram of a keypoint detection apparatus according to an embodiment of the present application, where the keypoint detection apparatus may include:
an image acquisition unit 10 for acquiring an image to be detected;
the thermodynamic diagram determining unit 20 is configured to process an image to be detected to obtain thermodynamic diagrams of a preset number, where each pixel in the thermodynamic diagrams is assigned with a score, and the score represents a probability that each pixel is a key point;
the key point information determining unit 30 is configured to determine, for each thermodynamic diagram, a key point according to the score assigned to each pixel point, and obtain corresponding key point information;
and the target key point determining unit 40 is configured to obtain a target key point by screening based on the information of each key point, and correct the coordinates of the target key point to obtain corrected coordinates of the target key point.
As can be seen from the foregoing technical solutions, in the device for detecting a key point provided in the embodiments of the present application, the thermodynamic diagram determining unit 20 processes the acquired image to be detected to obtain thermodynamic diagrams in a preset number, each pixel point in the thermodynamic diagrams is assigned with a score representing a probability that each pixel point is a key point, the key point information determining unit 30 determines a key point according to the score assigned to each pixel point for each thermodynamic diagram, and obtains corresponding key point information, and the target key point determining unit 40 obtains a target key point by screening based on each key point information, and corrects a coordinate of the target key point to obtain a corrected coordinate of the target key point.
Further, when detecting the key points, the thermodynamic diagram determining unit 20 may process the image to be detected to obtain a preset number of thermodynamic diagrams, and the key point information determining unit 30 determines a key point for each thermodynamic diagram to obtain a preset number of key points, but when the image to be detected only includes a part of information, for example, when only half of the face exists in the image to be detected, since another part of information does not actually exist in the image to be detected, the key points predicted for another part of information should not exist in the image to be detected, in the present application, the key point information corresponding to the key points is obtained by the key point information determining unit 30, and the key points that should not exist in the image to be detected can be screened out by the target key point determining unit 40 using the key point information, and correcting the coordinates of the target key points to obtain corrected coordinates of the target key points.
Optionally, the determining unit 30 of the key point information executes a process of determining a key point according to the score assigned to each pixel point, which may include:
and taking the pixel point with the maximum score in each thermodynamic diagram as a key point.
Optionally, the steps executed by the thermodynamic diagram determining unit, the key point information determining unit, and the target key point determining unit are completed by using a pre-trained target detection model, where the target detection model is obtained by training using a training image labeled with target key point coordinates as training data, and the key point detecting device may further include:
a target detection model training unit, configured to train to obtain the target detection model, where the target detection model may include: the method comprises the following steps that an input layer, a thermodynamic diagram acquisition layer, a key point information determination layer and a key point screening layer are sequentially cascaded;
the training process of the target detection model may include:
acquiring a training image through an input layer;
processing the training images through the thermodynamic diagram acquisition layer to obtain thermodynamic diagrams with preset quantity;
determining a key point according to the value distributed by each pixel point and aiming at each thermodynamic diagram through a key point information determination layer, and obtaining corresponding key point information;
screening to obtain target key points through a key point screening layer based on the information of each key point, and correcting the coordinates of the target key points to obtain corrected coordinates of the target key points;
and updating parameters of a target detection model by taking the corrected target key point coordinates approaching to the corresponding target key point coordinate labels in the training image as training targets.
The embodiment of the present application further provides a key point detecting apparatus, fig. 4 shows a hardware structure block diagram of the key point detecting apparatus, referring to fig. 4, the hardware structure of the key point detecting apparatus may include: at least one processor 1, at least one communication interface 2, at least one memory 3 and at least one communication bus 4;
in the embodiment of the application, the number of the processor 1, the communication interface 2, the memory 3 and the communication bus 4 is at least one, and the processor 1, the communication interface 2 and the memory 3 complete mutual communication through the communication bus 4;
the processor 1 may be a central processing unit CPU, or an application Specific Integrated circuit asic, or one or more Integrated circuits configured to implement embodiments of the present invention, etc.;
the memory 3 may include a high-speed RAM memory, and may further include a non-volatile memory (non-volatile memory) or the like, such as at least one disk memory;
wherein the memory stores a program and the processor can call the program stored in the memory, the program for: and realizing each processing flow in the key point detection method.
Embodiments of the present application further provide a storage medium, where a program suitable for execution by a processor may be stored, where the program is configured to: and realizing each processing flow in the key point detection method.
Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, the embodiments can be combined with each other, and the same and similar parts can be referred to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A method for detecting a keypoint, comprising:
acquiring an image to be detected;
processing an image to be detected to obtain thermodynamic diagrams with preset quantity, wherein each pixel point in the thermodynamic diagrams is distributed with a score, and the score represents the probability that each pixel point is a key point;
determining a key point according to the value distributed by each pixel point aiming at each thermodynamic diagram, and obtaining corresponding key point information;
and screening to obtain target key points based on the information of each key point, and correcting the coordinates of the target key points to obtain corrected coordinates of the target key points.
2. The method of claim 1, wherein determining a key point according to the score assigned to each pixel point comprises:
and taking the pixel point with the maximum score in each thermodynamic diagram as a key point.
3. The method according to claim 1 or 2, wherein the step of processing the image to be detected until the coordinates of the key points of the target are obtained by screening is completed by using a pre-trained target detection model;
the target detection model is obtained by training with a training image marked with target key point coordinates as training data.
4. The method of claim 3, wherein the object detection model comprises: the method comprises the following steps that an input layer, a thermodynamic diagram acquisition layer, a key point information determination layer and a key point screening layer are sequentially cascaded;
the training process of the target detection model comprises the following steps:
acquiring a training image through an input layer;
processing the training images through the thermodynamic diagram acquisition layer to obtain thermodynamic diagrams with preset quantity;
determining a key point according to the value distributed by each pixel point and aiming at each thermodynamic diagram through a key point information determination layer, and obtaining corresponding key point information;
screening to obtain target key points through a key point screening layer based on the information of each key point, and correcting the coordinates of the target key points to obtain corrected coordinates of the target key points;
and updating parameters of a target detection model by taking the corrected target key point coordinates approaching to the corresponding target key point coordinate labels in the training image as training targets.
5. The method according to claim 1 or 2, wherein the key point information comprises: keypoint coordinates and keypoint scores.
6. A keypoint detection device, comprising:
the image acquisition unit is used for acquiring an image to be detected;
the thermodynamic diagram determining unit is used for processing the image to be detected to obtain thermodynamic diagrams with preset quantity, wherein each pixel point in the thermodynamic diagrams is distributed with a score, and the score represents the probability that each pixel point is a key point;
the key point information determining unit is used for determining a key point according to the value distributed by each pixel point aiming at each thermodynamic diagram and obtaining corresponding key point information;
and the target key point determining unit is used for screening to obtain target key points based on the information of each key point, and correcting the coordinates of the target key points to obtain corrected coordinates of the target key points.
7. The apparatus according to claim 6, wherein the key point information determining unit performs a process of determining a key point according to the score assigned to each pixel point, including:
and taking the pixel point with the maximum score in each thermodynamic diagram as a key point.
8. The apparatus according to claim 6, wherein the thermodynamic diagram determining unit, the key point information determining unit, and the target key point determining unit perform the steps performed by using a pre-trained target detection model, wherein the target detection model is trained using a training image labeled with target key point coordinates as training data, and the key point detecting apparatus further comprises:
the target detection model training unit is used for training to obtain the target detection model, wherein the target detection model comprises:
the method comprises the following steps that an input layer, a thermodynamic diagram acquisition layer, a key point information determination layer and a key point screening layer are sequentially cascaded;
the training process of the target detection model comprises the following steps:
acquiring a training image through an input layer;
processing the training images through the thermodynamic diagram acquisition layer to obtain thermodynamic diagrams with preset quantity;
determining a key point according to the value distributed by each pixel point and aiming at each thermodynamic diagram through a key point information determination layer, and obtaining corresponding key point information;
screening to obtain target key points through a key point screening layer based on the information of each key point, and correcting the coordinates of the target key points to obtain corrected coordinates of the target key points;
and updating parameters of a target detection model by taking the corrected target key point coordinates approaching to the corresponding target key point coordinate labels in the training image as training targets.
9. A keypoint detection apparatus, comprising: a memory and a processor;
the memory is used for storing programs;
the processor, executing the program, performs the steps of the keypoint detection method of any of claims 1 to 5.
10. A readable storage medium having stored thereon a computer program, characterized in that the computer program, when being executed by a processor, carries out the steps of the keypoint detection method according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111371392.9A CN114067406A (en) | 2021-11-18 | 2021-11-18 | Key point detection method, device, equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111371392.9A CN114067406A (en) | 2021-11-18 | 2021-11-18 | Key point detection method, device, equipment and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114067406A true CN114067406A (en) | 2022-02-18 |
Family
ID=80278076
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111371392.9A Pending CN114067406A (en) | 2021-11-18 | 2021-11-18 | Key point detection method, device, equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114067406A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114998424A (en) * | 2022-08-04 | 2022-09-02 | 中国第一汽车股份有限公司 | Vehicle window position determining method and device and vehicle |
| CN117523645A (en) * | 2024-01-08 | 2024-02-06 | 深圳市宗匠科技有限公司 | Face key point detection method and device, electronic equipment and storage medium |
-
2021
- 2021-11-18 CN CN202111371392.9A patent/CN114067406A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114998424A (en) * | 2022-08-04 | 2022-09-02 | 中国第一汽车股份有限公司 | Vehicle window position determining method and device and vehicle |
| CN117523645A (en) * | 2024-01-08 | 2024-02-06 | 深圳市宗匠科技有限公司 | Face key point detection method and device, electronic equipment and storage medium |
| CN117523645B (en) * | 2024-01-08 | 2024-03-22 | 深圳市宗匠科技有限公司 | Face key point detection method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108399367B (en) | Hand motion recognition method and device, computer equipment and readable storage medium | |
| CN109657533B (en) | Pedestrian re-identification method and related product | |
| CN112488064B (en) | Face tracking method, system, terminal and storage medium | |
| CN108205654B (en) | Action detection method and device based on video | |
| WO2018028546A1 (en) | Key point positioning method, terminal, and computer storage medium | |
| CN111753643B (en) | Character gesture recognition method, character gesture recognition device, computer device and storage medium | |
| US20190244008A1 (en) | Manufacturing Part Identification Using Computer Vision And Machine Learning | |
| CN114067406A (en) | Key point detection method, device, equipment and readable storage medium | |
| WO2023010758A1 (en) | Action detection method and apparatus, and terminal device and storage medium | |
| CN111626123A (en) | Video data processing method and device, computer equipment and storage medium | |
| CN109960962B (en) | Image recognition method and device, electronic equipment and readable storage medium | |
| CN110969045B (en) | Behavior detection method and device, electronic equipment and storage medium | |
| CN112926462B (en) | Training method and device, action recognition method and device and electronic equipment | |
| CN108986137B (en) | Human body tracking method, device and equipment | |
| CN111444850A (en) | Picture detection method and related device | |
| CN112417970A (en) | Target object identification method, device and electronic system | |
| CN113160231A (en) | Sample generation method, sample generation device and electronic equipment | |
| CN111680546A (en) | Attention detection method, attention detection device, electronic equipment and storage medium | |
| CN112541394A (en) | Black eye and rhinitis identification method, system and computer medium | |
| CN110929555B (en) | Face recognition method and electronic device using same | |
| CN111353429A (en) | Interest degree method and system based on eyeball turning | |
| CN111507244B (en) | BMI detection method and device and electronic equipment | |
| CN112766065A (en) | Mobile terminal examinee identity authentication method, device, terminal and storage medium | |
| WO2022247118A1 (en) | Pushing method, pushing apparatus and electronic device | |
| CN110807403B (en) | User identity identification method and device and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |