Gesture guidance information generation method and device, equipment, storage medium and program product

Abstract

The invention provides a posture guidance information generation method and device, equipment, a storage medium and a program product, and relates to the technical field of computers. A specific embodiment of the method comprises the steps of determining a fatigue key point of a target object based on an image sequence containing the target object; generating posture prompt information based on the fatigue key points; and generating a target image for guiding the target object to adjust the posture based on the posture prompt information. The posture adjustment guidance effect provided by the embodiment is visual, and the method can be suitable for monitoring and adjusting the posture of the target object in various scenes.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technologies, and in particular, to a method and apparatus, a device, a storage medium, and a program product for generating pose guidance information. Background Art

[0002] With the development of technology and the enrichment of technological products, people's living patterns have become increasingly monotonous. For example, sitting in front of a computer for a long time to work, sitting at a desk for a long time to study, watching TV on a sofa, or chatting with friends on a mobile phone, etc. A long-term single living pattern is likely to cause health problems in the body. For example, muscle soreness induced by the same posture for a long time.

[0003] Existing pose detection and guidance methods based on image recognition, such as bad sitting posture detection, wrong fitness movement detection, etc., still have many limitations in practical applications and need to be further improved. Summary of the Invention

[0004] Embodiments of the present disclosure provide a method and apparatus, a device, a storage medium, and a program product for generating pose guidance information.

[0005] According to a first aspect, an embodiment of the present disclosure provides a method for generating a pose guidance image, including: determining fatigue key points of a target object based on an image sequence including the target object; generating pose prompt information based on the fatigue key points; and generating a target image for guiding the target object to adjust the pose based on the pose prompt information.

[0006] According to a second aspect, an embodiment of the present disclosure provides a device for generating a pose guidance image, including: a fatigue detection module configured to determine fatigue key points of the target object based on an image sequence including the target object; a pose adjustment module configured to generate pose prompt information based on the fatigue key points; and an image generation module configured to generate a target image for guiding the target object to adjust the pose based on the pose prompt information.

[0007] According to a third aspect, an embodiment of the present disclosure provides an electronic device, which includes one or more processors; a storage device having stored thereon one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method described in the first aspect.

[0008] According to a fourth aspect, an embodiment of the present disclosure provides a computer-readable medium having stored thereon a computer program, and when the program is executed by a processor, the method described in the first aspect is implemented.

[0009] According to a fifth aspect, embodiments of the present disclosure provide a computer program product, including a computer program which, when executed by a processor, implements the method described in the first aspect.

[0010] The method, device, equipment, storage medium and program product for generating pose guidance information according to the embodiments of the present disclosure can determine the fatigue key points of a target object based on an image sequence including the target object, generate pose prompt information based on the fatigue key points, and generate a target image for guiding the target object to adjust the pose. Through the monitoring of the target object's pose, local muscle fatigue of the target object can be detected, relaxation pose prompts for improving muscle fatigue can be given, and relaxation pose guidance images can be generated. The guidance effect of pose adjustment is intuitive and can be applied to the monitoring and adjustment of the target object's pose in various scenarios.

[0011] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become easily understandable through the following description. Description of the Drawings

[0012] Figure 1 is an exemplary system architecture diagram of an embodiment applying the pose guidance image generation method of the present disclosure; Figure 2 is a flowchart of the pose guidance image generation method according to an embodiment of the present disclosure; Figure 3 is a flowchart of determining the fatigue key points of a target object according to some embodiments of the present disclosure; Figure 4 is a schematic diagram of training a key point detection model according to some embodiments of the present disclosure; Figure 5 is a schematic diagram of training a muscle fatigue analysis model according to some embodiments of the present disclosure; Figure 6 is a flowchart of generating pose prompt information according to some embodiments of the present disclosure; Figure 7 is a schematic diagram of generating key point position modification information and target action generation information according to some embodiments of the present disclosure; Figure 8 is a schematic diagram of training an artificial intelligence generated content model according to some embodiments of the present disclosure; Figure 9 is a schematic diagram of a specific embodiment of the pose guidance image generation method according to the present disclosure; Figure 10 is a schematic diagram of an application scenario using the pose guidance image generation method of the present disclosure; Figure 11 is a schematic diagram of another application scenario using the posture-guided image generation method disclosed herein; Figure 12 is a schematic diagram of another application scenario using the posture-guided image generation method disclosed herein; Figure 13 is a schematic diagram of an embodiment of a posture guidance image generating device according to the present disclosure; Figure 14 It is a schematic structural diagram of an electronic device suitable for implementing the embodiments of the present disclosure. DETAILED DESCRIPTION

[0013] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0014] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in the present disclosure can be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0015] In the technical solutions disclosed herein, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0016] Figure 1 An exemplary system architecture 100 is shown to which embodiments of the gesture-guided image generation method of the present disclosure may be applied.

[0017] like Figure 1 As shown, system architecture 100 may include terminal devices 101, 102, and 103, a network 104, and a server 105. Network 104 is a medium for providing communication links between terminal devices 101, 102, and 103 and server 105, and between terminal devices. Network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0018] Users can use terminal devices 101, 102, 103 to interact with other terminal devices or server 105 via network 104 to receive or send messages, etc. Terminal devices 101, 102, 103 can be installed with client application software, such as image and video playback application software, communication application software, etc.

[0019] The terminal devices 101, 102, and 103 can be hardware or software. When the terminal devices 101, 102, and 103 are hardware, they can be various electronic devices, including but not limited to televisions, laptop computers, tablet computers, mobile phones, and projection devices. When the terminal devices 101, 102, and 103 are software, they can be installed in the above-listed electronic devices. They can be implemented as multiple software or software modules, or as a single software or software module. Specific limitations are not provided here.

[0020] The server 105 can be a server that provides various services. For example, based on an image sequence including a target object, determine the fatigue key points of the target object, generate pose prompt information based on the fatigue key points, and generate a target image for guiding the target object to adjust the pose based on the pose prompt information.

[0021] It should be noted that the server 105 can be hardware or software. When the server 105 is hardware, it can be implemented as a distributed server cluster composed of multiple servers or as a single server. When the server is software, it can be implemented as multiple software or software modules (such as those used to provide picture and video playback services), or as a single software or software module. Specific limitations are not provided here.

[0022] It should be pointed out that the method for generating a pose guidance image provided by the embodiments of the present disclosure can be executed by the server 105, or by the terminal devices 101, 102, and 103, or by the server 105 and the terminal devices 101, 102, and 103 in cooperation with each other. Correspondingly, each part (such as a module or a sub-module) included in the pose guidance image generation device can be entirely set in the server 105, or entirely set in the terminal devices 101, 102, and 103, or separately set in the server 105 and the terminal devices 101, 102, and 103.

[0023] It should be understood that Figure 1 the numbers of the terminal devices, the network, and the server in

[0024] Figure 2 show the flow 200 of the method for generating a pose guidance image according to the embodiments of the present disclosure. As Figure 2 shown, the flow 200 can include the following steps: Step 201, based on an image sequence including a target object, determine the fatigue key points of the target object.

[0025] In this embodiment, the execution subject (for example, Figure 1The server 105 or the terminal devices 101, 102, 103) therein can determine the fatigue key points of the target object according to the image sequence including the target object. Among them, the image sequence including the target object can be obtained by a camera collecting images of the target object over a period of time. Among them, the camera can be a camera configured on the terminal device, such as the front camera of a mobile phone. The camera can also be a camera communicatively connected to the terminal device, such as a camera communicatively connected to a mobile phone. The embodiments of the present disclosure do not limit this. The image sequence can include multiple frames of images. The multiple frames of images can be photos taken by the camera multiple times, or can also be multiple image frames in a video taken by the camera. The embodiments of the present disclosure do not limit this.

[0026] In this embodiment, the execution entity can obtain the image sequence including the target object locally or through the network, analyze the image sequence including the target object through a preset algorithm, determine the part of the target object where muscle fatigue occurs, and determine the human key points corresponding to this part to obtain the fatigue key points of the target object. For example, the preset algorithm can be an artificial intelligence model. The embodiments of the present disclosure do not limit this. The execution entity can also obtain the image sequence including the target object locally or through the network, analyze the image sequence including the target object according to a preset rule, determine the part of the target object where muscle fatigue occurs, and determine the human key points corresponding to this part to obtain the fatigue key points of the target object. For example, the preset rule can be set according to a knowledge base, etc. The embodiments of the present disclosure do not limit this.

[0027] Among them, the human key points refer to the feature points of some important parts of the human body, which are the human skeleton key points identified through computer vision technology and can be used to describe the posture of the human body. For example, the human key points can include the center point of the head, the joints of the shoulders, the joints of the hips, etc. The embodiments of the present disclosure do not limit the implementation method for the execution entity to determine the fatigue key points of the target object according to the image sequence including the target object.

[0028] Step 202, generate posture prompt information based on the fatigue key points.

[0029] In this embodiment, the above-mentioned execution entity can generate pose prompt information according to the fatigue key points. Among them, the pose prompt information can include adjustment information for the current pose of the target object. The pose prompt information can include local adjustment information for the current pose of the target object, and by changing the local pose of the target object, the purpose of relaxing the part where muscle fatigue occurs can be achieved. For example, information for changing the position of the fatigue key points can be generated based on the current pose of the target object. The pose prompt information can also include overall adjustment information for the current pose of the target object, and by making the target object make a specific pose different from the current pose, the purpose of relaxing the part where muscle fatigue occurs can be achieved. For example, information for generating a new pose unrelated to the current pose of the target object can be generated. The embodiments of the present disclosure do not limit the form of the pose prompt information.

[0030] In this embodiment, the execution entity can analyze the position of the fatigue key points through a preset algorithm, and combine the image sequence including the target object to generate pose prompt information for adjusting the pose of the target object. For example, the preset algorithm can be an artificial intelligence model, and the embodiments of the present disclosure do not limit this. The execution entity can also analyze the position of the fatigue key points according to a preset rule, and combine the image sequence including the target object to generate pose prompt information for adjusting the pose of the target object. For example, the preset rule can be set according to a knowledge base, etc., and the embodiments of the present disclosure do not limit this. The embodiments of the present disclosure do not limit the implementation method for the execution entity to generate pose prompt information according to the fatigue key points.

[0031] Step 203: Generate a target image for guiding the target object to adjust the pose based on the pose prompt information.

[0032] In this embodiment, the above-mentioned execution entity can generate a target image for guiding the target object to adjust the pose according to the pose prompt information. Among them, the target image can be a guiding image presenting the relaxed pose after the pose adjustment, and the target object can adjust the current pose according to the relaxed pose presented in the target image to relax the part where muscle fatigue occurs. Among them, the target image can be a guiding picture presenting the relaxed pose. For example, when the relaxed pose is a static pose, a guiding picture presenting the relaxed pose can be generated according to the pose prompt information. The target image can also be a guiding video presenting the relaxed pose. For example, when the relaxed pose is a dynamic pose, a guiding video presenting the relaxed pose can be generated according to the pose prompt information. The embodiments of the present disclosure do not limit the form of the target image.

[0033] In this embodiment, the execution entity may generate a target image for guiding the target object to adjust the current posture according to the posture prompt information through a preset algorithm. For example, the preset algorithm may be an artificial intelligence model, and the embodiments of the present disclosure do not limit this. The execution entity may also generate a target image for guiding the target object to adjust the current posture according to the posture prompt information according to a preset rule. For example, the preset rule may be set according to a knowledge base, etc., and the embodiments of the present disclosure do not limit this. The embodiments of the present disclosure do not limit the implementation method for the execution entity to generate a target image for guiding the target object to adjust the posture according to the posture prompt information.

[0034] The posture guidance image generation method provided by the embodiments of the present disclosure determines the fatigue key points of the target object based on an image sequence including the target object, generates posture prompt information based on the fatigue key points, and generates a target image for guiding the target object to adjust the posture. It can detect the local muscle fatigue of the target object through the monitoring of the target object's posture, give relaxation posture prompts for improving muscle fatigue, and generate relaxation posture guidance images. The guidance effect of posture adjustment is intuitive and can be applied to the monitoring and adjustment of the target object's posture in various scenarios without being limited to a specific scenario. For example, the applicable scenarios may include scenarios such as sitting for a long time, standing for a long time, running, and walking slowly, and there is no need to preset the application scenario in advance and distinguish specific scenarios for posture monitoring and adjustment.

[0035] Figure 3 Flowchart 300 for determining the fatigue key points of the target object in some embodiments of the present disclosure is shown. As Figure 3 shown, flowchart 300 may include the following steps: Step 301, perform key point detection on the target object in each image of the image sequence to obtain the initial key points of the image sequence.

[0036] In this embodiment, the execution entity (for example, Figure 1The server 105 or the terminal devices 101, 102, 103) in it can perform key point detection on the target object in each image of the image sequence to obtain the initial key points of the image sequence. Among them, the camera can collect images of the target object from any angle to obtain an image sequence containing the target object. In an optional example, the camera can collect images of the target object from the side of the target object, and the images in the image sequence are side images of the target object. In another optional example, the camera can collect images of the target object from the front of the target object, and the images in the image sequence are front images of the target object. The execution subject can perform human key point detection on the target object in each image of the image sequence through a preset algorithm or according to a preset rule to obtain the human key points of the target object in each image of the image sequence as the initial key points of the image sequence. The implementation method of performing key point detection on the image sequence in the embodiments of the present disclosure is not limited.

[0037] In an optional example, 17 initial key points of the target object in the image sequence can be obtained through key point detection, which may include 5 human key points of the head, 8 human key points of the torso and upper limbs, and 2 human key points of the lower limbs. In other optional examples of the present disclosure, the number of initial key points can also be more than 17, such as 25 human key points, or the number of initial key points can also be less than 17, such as 14 human key points. The embodiments of the present disclosure do not limit this.

[0038] In some optional embodiments, the key point detection model can be used to perform key point detection on the target object in each image of the image sequence to obtain the initial key points of the image sequence. Among them, an existing key point detection model can be used to perform key point detection on the image sequence, or the sample images obtained by data augmentation through an Artificial Intelligence Generated Content (AIGC) model can be used to train the existing key point detection model, and the trained key point detection model can be used to perform key point detection on the image sequence. The embodiments of the present disclosure do not limit this.

[0039] Optionally, in step 301, key points of the target object in each image of the image sequence are detected to obtain the initial key points of the image sequence, which may include: inputting the image sequence into a key point detection model to output the initial key points of the image sequence. Among them, the key point detection model is trained based on the sample images obtained by data augmentation using the first artificial intelligence-generated content model. In this embodiment, the AIGC model is used to perform data augmentation on the sample images, and the existing key point detection model is optimized according to the sample images obtained by data augmentation, which can improve the impact of insufficient data on the key point detection model, and can avoid the limitations caused by the shooting angle of the monocular camera, such as the differences in the key point detection results of the human body side and front, and can improve the accuracy of the existing key point detection model and improve the accuracy of human key point detection.

[0040] As Figure 4 shown, Figure 4 The figure shows the process of training the key point detection model in some embodiments of the present disclosure. Among them, diverse human pose images can be generated by inputting text to obtain an initial sample image dataset. For example, by inputting "generate a photo of a child doing homework, facing the camera" into the text-to-image model, the corresponding human pose image can be generated. Alternatively, a large number of unlabeled real human pose images can be collected as the initial sample images to obtain the initial sample image dataset. The embodiments of the present disclosure do not limit the manner of obtaining the initial sample image dataset. Then, the human pose images in the initial sample image dataset are input into the AIGC model to remove occluding objects, such as backpacks, tables, etc., and texture filling is performed on the missing areas in the image after removing the occluding objects. The obtained human pose images are used as new sample images to obtain a new sample image dataset. Then, the pre-trained key point detection model is used to perform human key point detection on the initial sample images and the new sample images respectively, and the key point detection results of the new sample images and the corresponding initial sample images are compared. A weight is assigned to each key point according to the difference in the key point detection results of the two, so that the closer the key point detection results are, the higher the weight of the key point. Therefore, the weight of each key point can also reflect the confidence of this key point.

[0041] Optionally, the training steps of the key point detection model may include: inputting the first sample image of the sample image dataset into the first artificial intelligence generated content model to output a second sample image with the occluder removed from the first sample image and texture filled; inputting the first sample image and the second sample image into the key point detection model respectively to output the human key points in the first sample image and the second sample image; setting weights for the human key points in the first sample image and the second sample image based on the differences between the human key points in the second sample image and the corresponding first sample image to obtain weighted key points; and fine-tuning the key point detection model based on the first sample image, the second sample image, and the weighted key points. In this embodiment, the AIGC model is used to remove the occluder from the sample image and fill the texture, and the existing key point detection model is optimized according to the filled sample image, which can improve the influence of occlusion on the key point detection model and avoid the limitations caused by the shooting angle of the monocular camera, such as the influence of the half-body or occluded situation of the human body on the key point detection result, improve the accuracy of the existing key point detection model, and improve the accuracy of human key point detection.

[0042] Step 302: Based on the initial key points of the image sequence and the personalized information of the target object, perform muscle fatigue analysis on the target object to determine the fatigue key points.

[0043] In this embodiment, the above execution subject may perform muscle fatigue analysis on the target object based on the initial key points of the image sequence and the personalized information of the target object to determine the fatigue key points. Among them, the personalized information may include the human body information of the target object, such as height, weight, chest circumference, waist circumference, etc. The execution subject may perform muscle fatigue analysis on the target object according to the initial key points of the image sequence and the personalized information of the target object through a preset algorithm or according to a preset rule, determine the part where the target object has muscle fatigue, and determine the human key points corresponding to this part to obtain the fatigue key points of the target object. The implementation method of muscle fatigue analysis according to the image sequence and personalized information in the present disclosure is not limited.

[0044] In some alternative embodiments, a muscle fatigue analysis model can be used to perform muscle fatigue analysis on a target object based on the initial key points of the image sequence and the personalized information of the target object, so as to determine the fatigue key points of the target object. Among them, the muscle fatigue analysis model can be trained based on a human body posture sample image sequence, the personalized information of the human body in the human body posture sample image sequence, and the muscle information of the human body that generates fatigue. Optionally, step 303 of performing muscle fatigue analysis on the target object based on the initial key points of the image sequence and the personalized information of the target object to determine the fatigue key points may include: inputting the initial key points of the image sequence and the personalized information of the target object into the muscle fatigue analysis model, and outputting the fatigue key points of the target object, where the muscle fatigue analysis model is trained based on a human body posture sample image sequence, the personalized information of the human body, and the muscle information that generates fatigue. In this embodiment, by using the muscle fatigue analysis model to perform muscle fatigue analysis on the target object in the image sequence, the muscle information of the target object that generates fatigue can be quickly obtained. Since the degree of muscle fatigue is not only related to the human body posture, but also affected by the response of the human body information, by adding the personalized information of the target object during the muscle fatigue analysis process, the muscle information of the target object that generates fatigue can be obtained more accurately.

[0045] As Figure 5 shown, Figure 5The figure shows the process of training a muscle fatigue analysis model according to some embodiments of the present disclosure. Among them, a large number of image sequences of different human postures can be collected as the human posture sample image sequences to obtain a sample image sequence data set. For example, image sequences of different human postures from multiple angles can be collected to obtain a sample image sequence data set, where different human postures can include different static postures, such as sitting, lying down, squatting, etc., and different dynamic postures, such as running, walking slowly, etc. It is also possible to collect a large number of unannotated human posture videos as the human posture sample image sequences to obtain a sample image sequence data set. For example, a large number of videos of humans doing fitness exercises can be collected as the human posture sample image sequences to obtain a sample image sequence data set. The embodiments of the present disclosure do not limit the manner of obtaining the sample image sequence data set. The electromyogram of the human body can be obtained by collecting the human muscle electrical signals corresponding to the human posture sample image sequences in the sample image sequence data set. By performing spectral analysis and amplitude analysis on the electromyogram of the human body, the muscle information of the human body that generates fatigue can be obtained. Combining with the human key points, the fatigue key points can be determined, thereby obtaining the muscle fatigue result. It is also possible to analyze the text description of the video or the action behavior of the human body in the video to obtain the muscle information of the human body that generates fatigue. Combining with the human key points, the fatigue key points can be determined, thereby obtaining the muscle fatigue result. For example, a neural network can be used to analyze the video description or video behavior to obtain the muscle information of a large number of videos of humans doing fitness exercises that generate fatigue. The personal information of the human body in the human posture sample image sequences in the sample image sequence data set can also be used as a training condition to train the muscle fatigue analysis model.

[0046] Optionally, the training steps of the muscle fatigue analysis model may include: inputting the human posture sample image sequences of the sample image sequence data set into a key point detection model to output the human key points in each image of the human posture sample image sequences; analyzing the human muscle electrical signals in the human posture sample image sequences or analyzing the actions of the humans in the human posture sample image sequences to determine the muscle information that generates fatigue; training the muscle fatigue analysis model based on the human key points, the personal information of the human body, and the muscle information that generates fatigue in the human posture sample image sequences. In this embodiment, by analyzing the human muscle electrical signals or human actions in the human posture sample image sequences, and combining the personal information of the human body, the muscle information of the human body that generates fatigue can be determined. Combining with the human key points, the fatigue key points of the human body can be determined as the muscle fatigue result, which can ensure the accuracy of the muscle fatigue analysis model and improve the accuracy of detecting the human fatigue key points.

[0047] In some alternative embodiments of the present disclosure, the pose prompt information in step 202 may include key point position modification information and target action generation information. The process 200 may further include: a step of determining the pose change frequency of the target object based on the initial key points of the image sequence. Figure 6 FIG. 600 shows a process of generating pose prompt information according to some embodiments of the present disclosure. As Figure 6 shown, the process 600 may include the following steps: Step 601, determining the pose change frequency of the target object based on the initial key points of the image sequence.

[0048] In this embodiment, the execution entity (e.g., Figure 1 the server 105 or the terminal devices 101, 102, 103 in FIG. 1) may determine the pose change frequency of the target object according to the initial key points of the image sequence. Among them, the execution entity may analyze the positions of the human key points of the target object in each image of the image sequence to obtain the pose change frequency of the target object in the image sequence. For example, it may be determined that the pose of the target object has changed according to the position change of the human key points of the target object in two adjacent frames exceeding a preset range. On the contrary, it may be determined that the pose of the target object has not changed according to the position change of the human key points of the target object in two adjacent frames within the preset range. The execution entity may obtain the pose change frequency of the target object by counting the number of times the pose of the target object changes in the image sequence in combination with the time length of the image sequence. The embodiments of the present disclosure do not limit the implementation method for the execution entity to determine the pose change frequency of the target object according to the initial key points of the image sequence.

[0049] Step 602, determining the pose type of the target object based on the pose change frequency, where the pose type includes a static pose and a dynamic pose.

[0050] In this embodiment, the above-mentioned execution entity may determine the pose type of the target object according to the pose change frequency, where the pose type may include a static pose and a dynamic pose. Among them, according to the pose change frequency of the target object in the image sequence, the pose of the target object in the image sequence may be divided into a static pose and a dynamic pose. A static pose may refer to a pose that does not change for a long time of the target object, such as a sitting or standing still pose for a long time. A dynamic pose may refer to a pose that changes in a short time of the target object, such as a running or walking slowly pose. For example, a frequency threshold may be preset, and the pose change frequency is compared with the preset frequency threshold. The pose of the target object with a pose change frequency greater than the preset frequency threshold is determined as a static pose, and the pose of the target object with a pose change frequency less than or equal to the preset frequency threshold is determined as a dynamic pose. The embodiments of the present disclosure do not limit the implementation method for the execution entity to determine the pose type of the target object according to the pose change frequency.

[0051] Step 603: In response to the posture type of the target object being a dynamic posture, generate target action generation information based on the fatigue key points.

[0052] In this embodiment, in response to the posture type of the target object being a dynamic posture, the above-mentioned execution entity can generate target action generation information according to the fatigue key points. Among them, for the dynamic posture of the target object, the execution entity can relax the fatigued muscles by making an overall adjustment to the posture of the target object and changing the overall posture of the target object. Among them, by changing the overall posture of the target object, a specific posture unrelated to the current posture of the target object for improving muscle fatigue can be obtained. For example, a stretching action for relaxing muscles after fitness can be obtained. The embodiment of the present disclosure does not limit the type of the specific posture. The execution entity can generate target action generation information according to the fatigue key points and obtain a specific posture different from the current posture of the target object.

[0053] Step 604: In response to the posture type of the target object being a static posture, generate key point position modification information based on the fatigue key points.

[0054] In this embodiment, in response to the posture type of the target object being a static posture, the above-mentioned execution entity can generate key point position modification information according to the fatigue key points. Among them, for the static posture of the target object, the execution entity can relax the fatigued muscles by making a partial adjustment to the posture of the target object and changing the partial posture of the target object. Among them, by changing the partial posture of the target object, the position of the key points related to muscle fatigue can be modified on the basis of the current posture of the target object. The key points related to muscle fatigue can include fatigue key points and can also include other key points other than fatigue key points. The embodiment of the present disclosure does not limit this. The execution entity can generate key point position modification information according to the fatigue key points and modify the positions of the key points related to muscle fatigue of the target object.

[0055] As Figure 7 shown, Figure 7The figure shows a process for generating key point position modification information and target action generation information for some embodiments of the present disclosure. Among them, the initial key points of the N-frame image sequence can be preprocessed to obtain the posture change frequency of the N-frame image sequence. According to the posture change frequency, the postures of the target object in the N-frame image sequence can be divided into static postures and dynamic postures. For dynamic postures, creative text prompt information can be generated based on the fatigue key points resulting from muscle fatigue. The creative text prompt information is the information used to generate the target action. For example, the creative text prompt information is "generate an image or video of relaxing the shoulders". For static postures, the duration of the static posture can be counted, and it can be determined whether to stop the analysis of the muscle fatigue analysis model according to the duration of the static posture, so as to reduce the time-consuming of the muscle fatigue analysis model for analysis. When the duration of the static posture is long, for example, when the duration of the static posture is greater than a preset time threshold, the muscle fatigue analysis model can be stopped from continuing the muscle fatigue analysis. The key point position modification information can be generated based on the initial muscle fatigue result obtained by the muscle fatigue analysis model and the initial key points of a frame of the N-frame image sequence serving as the reference key points. Among them, the key point position modification information can include: making a certain range of changes to the angles and positions of the key points related to muscle fatigue based on the fatigue key points. In this embodiment, by determining the posture change frequency of the target object and dividing the postures of the target object into dynamic postures and static postures according to the posture change frequency, it is convenient to generate different posture prompt information according to different posture types, which can meet the requirements of different application scenarios and provide a better user experience for users.

[0056] In some alternative embodiments of the present disclosure, step 203 of generating a target image for guiding the target object to adjust the posture based on the posture prompt information may include: inputting the posture prompt information and the reference image into the second artificial intelligence generation content model, and outputting a target image for guiding the target object to adjust the posture. Among them, the posture prompt information can be input into the second artificial intelligence generation content model as a control condition, and the modification of the reference image by the second artificial intelligence generation content model can be controlled through the control condition, so that the second artificial intelligence generation content model generates a target image that conforms to the posture prompt information by modifying the reference image. Among them, the reference image can be a frame of the image sequence containing the target object, or a frame of the image from the reference image database that has nothing to do with the image sequence containing the target object. The embodiments of the present disclosure do not limit this.

[0057] As Figure 8 shown, Figure 8The figure shows the process of training an artificial intelligence generated content model for some embodiments of the present disclosure. Among them, the second artificial intelligence generated content model may include a key point control module and a time domain control module. The key point control module is used to generate control conditions for modifying the positions of local key points, and the time domain control module is used to generate control conditions for images with continuity in the time dimension. The key point position modification information and the target action generation information can be respectively input into the key point control module and the time domain control module to generate corresponding control conditions, so as to control the reference image to generate target images that conform to the key point position modification information and the target action generation information respectively. When training the second artificial intelligence generated content model, the reference image uploaded by the user or the image captured by the camera can be input into the encoder of the second artificial intelligence generated content model. The encoder will add noise Zr to the image to generate a noisy image. The noisy image and the control conditions generated by the key point position modification information or the target action generation information are input into the decoder of the second artificial intelligence generated content model to control the decoding of the noisy image and generate a frame of image or video for guiding the target object to adjust its posture. Among them, the key point control module can be trained using an N-frame human video dataset, and the time domain control module can be trained using an N-frame muscle stretching video.

[0058] In some alternative examples, inputting the pose prompt information and the reference image into the second artificial intelligence generated content model to output a target image for guiding the target object to adjust its posture may include: in response to the pose type of the target object being a static pose, inputting a frame of image in the image sequence as the reference image into the encoder of the second artificial intelligence generated content model to output a noisy image; inputting the key point position modification information into the key point control module of the second artificial intelligence generated content model to output key point position control information; inputting the noisy image and the key point position control information into the decoder of the second artificial intelligence generated content model to output a frame of image for guiding the target object to adjust its posture. Among them, the reference image is an actual scene image.

[0059] In some other alternative examples, inputting the pose prompt information and the reference image into the second artificial intelligence generated content model to output a target image for guiding the target object to adjust its posture may include: in response to the pose type of the target object being a dynamic pose, obtaining a frame of image from the reference image database as the reference image and inputting it into the encoder of the second artificial intelligence generated content model to output a noisy image; inputting the target action generation information into the time domain control module of the second artificial intelligence generated content model to output action time domain control information; inputting the noisy image and the action time domain control information into the decoder of the second artificial intelligence generated content model to output a video for guiding the target object to adjust its posture. Among them, the reference image is a customized scene image.

[0060] In this embodiment, the AIGC model can use images from multiple sources as reference images, and generate various pose guidance images or videos according to the reference images, which can meet the requirements of different application scenarios and provide users with a better user experience.

[0061] As Figure 9 shown, Figure 9 Figure 1 shows a specific embodiment of the pose guidance image generation method of the present disclosure. Among them, the pose guidance image generation method can be divided into two stages. The first stage can include fatigue detection and pose correction. Among them, fatigue detection can include a key point detection model and a muscle fatigue analysis model, and pose correction can include a preprocessing module and a correction module. The second stage is used for generating guidance results and can include an AIGC model. Its specific process can include: In the first stage, N frames of images of the target object are captured from any angle by a monocular camera and input into the key point detection model as an image sequence. Key point detection is performed on the target object in each image of the image sequence to obtain N frames of reference key points. The N frames of reference key points are processed by the preprocessing module to determine the pose change frequency of the target object in the N frames of images. The personalized human data of the target object in the N frames of images and the N frames of reference key points are input into the muscle fatigue analysis model to perform muscle fatigue analysis on the target object to determine the muscle fatigue result. The muscle fatigue result is the fatigue key points of the target object. The pose change frequency of the target object, the muscle fatigue result, and the n N-1 th frame of reference key points selected from the N frames of reference key points are input into the correction module to generate corrected key point information or creative hint information. Among them, for static poses with a small pose change frequency of the target object, the correction module generates corrected key point information, and the corrected key point information is information for locally modifying the positions of the key points based on the current pose of the target object. For dynamic poses with a large pose change frequency of the target object, the correction module generates creative hint information, and the creative hint information is information for generating specific actions unrelated to the current pose of the target object.

[0062] In the second stage, the corrected key point information or creative hint information and the reference image are input into the AIGC model. The corrected key point information or creative hint information is used as a control condition to control the AIGC model to generate a target image for guiding the target object to adjust the pose according to the reference image. Among them, the reference image can be the n N -1 th frame selected from the N frames of images, or a frame of image selected from the reference image database. One frame of image for guiding the target object to adjust the pose can be generated according to the corrected key point information. A video for guiding the target object to adjust the pose can be generated according to the creative hint information.

[0063] As Figure 10As shown, Figure 10 Fig. Figure 10 shows an application scenario of using the posture guidance image generation method of the present disclosure. This application scenario is a scenario where a user watches TV. Among them, line c divides Figure 10 into two regions. Region A above is a solution using the prior art, and region B below is a solution using the posture guidance image generation method of the present disclosure. According to the solution of region A, when an abnormal sitting posture of the user is detected, the user is reminded of the abnormal sitting posture by voice, or the correct sitting posture is informed to the user. According to the solution of region B, the muscle fatigue of the user can be monitored for a long time. When the muscle fatigue of the user's sedentary arm is detected, a picture for guiding the user to adjust the arm posture and relax the arm muscles can be generated, which can facilitate the user to correctly adjust the posture according to the guiding picture.

[0064] As Figure 11 shown, Figure 11 Fig. Figure 11 shows another application scenario of using the posture guidance image generation method of the present disclosure. This application scenario is a scenario where a user uses a smart display. Similar to Figure 10 the above, line c divides Figure 11 into two regions. Region A above is a solution using the prior art, and region B below is a solution using the posture guidance image generation method of the present disclosure. According to the solution of region A, when an abnormal sitting posture of the user is detected, the user is reminded of the abnormal sitting posture by voice, or the correct sitting posture is informed to the user. According to the solution of region B, the muscle fatigue of the user can be monitored for a long time. When the muscle fatigue of the user's sedentary neck is detected, a picture for guiding the user to adjust the neck posture and relax the neck muscles can be generated, which can facilitate the user to correctly adjust the posture according to the guiding picture.

[0065] As Figure 12 shown, Figure 12 Fig. Figure 12 shows yet another application scenario of using the posture guidance image generation method of the present disclosure. This application scenario is a scenario where a user does fitness exercises. The posture guidance image generation method of the present disclosure can monitor the muscle fatigue of the user for a long time during the fitness process of the user. When the muscle fatigue of the user's legs caused by squats is detected, a video for guiding the user to stretch the legs and relax the leg muscles can be generated, which can facilitate the user to correctly make muscle relaxation movements according to the guiding video.

[0066] Further referring to Figure 13 , as an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a posture guidance image generation device. This device embodiment corresponds to the method embodiment shown in Figure 2 , and this device can be specifically applied to various electronic devices.

[0067] As Figure 13As shown in the figure, the posture guidance image generation device of this embodiment may include: a fatigue detection module 1301, a posture adjustment module 1302, and an image generation module 1303.

[0068] Among them, the fatigue detection module 1301 may be configured to determine the fatigue key points of the target object based on an image sequence including the target object.

[0069] The posture adjustment module 1302 may be configured to generate posture prompt information based on the fatigue key points.

[0070] The image generation module 1303 may be configured to generate a target image for guiding the target object to adjust the posture based on the posture prompt information.

[0071] In some alternative implementation manners of the present disclosure, the fatigue detection module 1301 may include: A key point detection sub-module, which may be configured to perform key point detection on the target object in each image of the image sequence to obtain the initial key points of the image sequence; A muscle fatigue analysis sub-module, which may be configured to perform muscle fatigue analysis on the target object based on the initial key points of the image sequence and the personalized information of the target object to determine the fatigue key points.

[0072] In some alternative implementation manners of the present disclosure, the key point detection sub-module is further configured to: Input the image sequence into a key point detection model, and output the initial key points of the image sequence, where the key point detection model is trained based on sample images obtained by data augmentation using a first artificial intelligence content generation model.

[0073] In some alternative implementation manners of the present disclosure, the training steps of the key point detection model may include: Input a first sample image of a sample image dataset into the first artificial intelligence content generation model, and output a second sample image with the occluders removed from the first sample image and texture filled; Input the first sample image and the second sample image into the key point detection model respectively, and output the human key points in the first sample image and the second sample image respectively; Based on the difference between the human key points in the second sample image and the corresponding first sample image, set weights for the human key points in the first sample image and the second sample image to obtain weighted key points; Fine-tune the key point detection model based on the first sample image, the second sample image, and the weighted key points.

[0074] In some alternative implementation manners of the present disclosure, the muscle fatigue analysis sub-module is further configured to: Input the initial key points of the image sequence and the personalized information of the target object into a muscle fatigue analysis model, and output the fatigue key points of the target object, where the muscle fatigue analysis model is trained based on a human body posture sample image sequence, the personalized information of the human body, and the muscle information generating fatigue.

[0075] In some alternative implementation manners of the present disclosure, the training steps of the muscle fatigue analysis model may include: Input the human body posture sample image sequence of the sample image sequence dataset into a key point detection model, and output the human key points in each image of the human body posture sample image sequence; Analyze the electromyogram signals of the human body in the human body posture sample image sequence or analyze the actions of the human body in the human body posture sample image sequence, and determine the muscle information generating fatigue; Train the muscle fatigue analysis model based on the human key points of the human body posture sample image sequence, the personalized information of the human body, and the muscle information generating fatigue.

[0076] In some alternative implementation manners of the present disclosure, the posture prompt information includes key point position modification information and target action generation information, and the device further includes: A frequency detection module, which may be configured to determine the posture change frequency of the target object based on the initial key points of the image sequence; The posture adjustment module 1302 may include: A posture classification sub-module, which may be configured to determine the posture type of the target object based on the posture change frequency, where the posture type includes a static posture and a dynamic posture; A first information generation sub-module, which may be configured to generate the target action generation information based on the fatigue key points in response to the posture type of the target object being a dynamic posture; A second information generation sub-module, which may be configured to generate the key point position modification information based on the fatigue key points in response to the posture type of the target object being a static posture.

[0077] In some alternative implementation manners of the present disclosure, the image generation module 1303 is further configured to: Input the posture prompt information and a reference image into a second artificial intelligence generation content model, and output a target image for guiding the target object to adjust the posture.

[0078] In some alternative implementation manners of the present disclosure, the image generation module 1303 is further configured to: In response to the pose type of the target object being a static pose, one frame of the image sequence is used as the reference image and input into the encoder of the second artificial intelligence generated content model, and an image with added noise is output; The key point position modification information is input into the key point control module of the second artificial intelligence generated content model, and key point position control information is output; The image with added noise and the key point position control information are input into the decoder of the second artificial intelligence generated content model, and one frame of the image for guiding the target object to adjust the pose is output.

[0079] In some alternative implementation manners of the present disclosure, the image generation module 1303 is further configured to: In response to the pose type of the target object being a dynamic pose, one frame of the image is obtained from the reference image database as the reference image and input into the encoder of the second artificial intelligence generated content model, and an image with added noise is output; The target action generation information is input into the time domain control module of the second artificial intelligence generated content model, and action time domain control information is output; The image with added noise and the action time domain control information are input into the decoder of the second artificial intelligence generated content model, and a video for guiding the target object to adjust the pose is output.

[0080] According to an embodiment of the present disclosure, the present disclosure further provides an electronic device, a computer-readable storage medium, and a computer program product.

[0081] Figure 14 is a block diagram of an electronic device suitable for implementing the embodiments of the present disclosure. As Figure 14 shown, the electronic device includes: one or more processors 1401, a memory 1402, and an interface for connecting each component, including a high-speed interface and a low-speed interface. Each component is interconnected using different buses and can be installed on a common motherboard or installed in other ways as needed. The processor can process instructions executed within the electronic device, including instructions stored in the memory or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In other embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple electronic devices can be connected, and each device provides some necessary operations (for example, as a server array, a set of blade servers, or a multi-processor system). Figure 14 In the figure, one processor 1401 is taken as an example.

[0082] Memory 1402 is the non-transitory computer-readable storage medium provided by the present disclosure. Among them, the memory stores instructions executable by at least one processor, so that the at least one processor executes the posture guidance image generation method provided by the present disclosure. The non-transitory computer-readable storage medium of the present disclosure stores computer instructions, and these computer instructions are used to cause a computer to execute the posture guidance image generation method provided by the present disclosure.

[0083] As a non-transitory computer-readable storage medium, memory 1402 can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the posture guidance image generation method in the embodiments of the present disclosure (for example, the Figure 13 fatigue detection module 1301, posture adjustment module 1302, and image generation module 1303 shown in the appendix). By running the non-transitory software programs, instructions, and modules stored in memory 1402, processor 1401 executes various functional applications and data processing of the server, that is, implements the posture guidance image generation method in the above method embodiments.

[0084] Memory 1402 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created by the use of the electronic device for historical playback videos, etc. In addition, memory 1402 may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage devices. In some embodiments, memory 1402 may optionally include a memory remotely set relative to processor 1401, and these remote memories can be connected to the electronic device that executes the posture guidance image generation method through a network. Examples of the above network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and combinations thereof.

[0085] The electronic device for the posture guidance image generation method may further include: an input device 1403 and an output device 1404. Processor 1401, memory 1402, input device 1403, and output device 1404 can be connected through a bus or other means, Figure 14 taking the connection through the bus as an example.

[0086] The input device 1403 can receive input digital or character information, as well as key signal inputs related to user settings and function controls of an electronic device that plays video streams, such as input devices like touchscreens, keypads, mice, trackpads, touchpads, pointing sticks, one or more mouse buttons, trackballs, joysticks, etc. The output device 1404 can include a display device, an auxiliary lighting device (e.g., LED), and a haptic feedback device (e.g., a vibration motor), etc. The display device can include, but is not limited to, a liquid crystal display (LCD), a light-emitting diode (LED) display, and a plasma display. In some embodiments, the display device can be a touchscreen.

[0087] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuitry, integrated circuit systems, application specific ASICs (application specific integrated circuits), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special or general programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0088] These computing programs (also referred to as programs, software, software applications, or code) include machine instructions for a programmable processor and can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms “machine-readable medium” and “computer-readable medium” refer to any computer program product, device, and / or apparatus (e.g., a disk, an optical disk, a memory, a programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term “machine-readable signal” refers to any signal for providing machine instructions and / or data to a programmable processor.

[0089] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and a pointing device (e.g., a mouse or a trackball) through which the user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, speech input, or tactile input).

[0090] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), and the Internet.

[0091] A computer system can include a client and a server. The client and the server are generally far from each other and usually interact through a communication network. The client - server relationship is created by computer programs running on the respective computers and having a client - server relationship with each other.

[0092] According to the technical solution of the embodiment of the present disclosure, the accuracy and effectiveness of fault location by the trained fault location model are effectively improved.

[0093] It should be understood that various forms of the processes shown above can be used, re - ordering, adding, or deleting steps. For example, the steps recited in this application can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution disclosed in the present disclosure can be achieved, and no limitation is made herein.

[0094] The above - mentioned specific embodiments do not constitute a limitation to the protection scope of the present disclosure. Those skilled in the art should understand that various modifications, combinations, sub - combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present disclosure shall be included within the protection scope of the present disclosure.

Claims

1. A method for generating a posture guidance image, comprising: Determining fatigue key points of a target object based on an image sequence including the target object; Generating posture prompt information based on the fatigue key points; Generating a target image for guiding the target object to adjust the posture based on the posture prompt information.

2. The method according to claim 1, wherein, The determining the fatigue key points of the target object based on the image sequence including the target object includes: Performing key point detection on the target object in each image of the image sequence to obtain initial key points of the image sequence; Performing muscle fatigue analysis on the target object based on the initial key points of the image sequence and the personalized information of the target object to determine the fatigue key points.

3. The method according to claim 2, wherein, The performing key point detection on the target object in each image of the image sequence to obtain the initial key points of the image sequence includes: Inputting the image sequence into a key point detection model to output the initial key points of the image sequence, where the key point detection model is trained with sample images obtained by data augmentation based on a first artificial intelligence generated content model.

4. The method according to claim 3, wherein The training steps of the key point detection model include: Inputting a first sample image of a sample image data set into the first artificial intelligence generated content model to output a second sample image with the occluder removed from the first sample image and texture filled; Inputting the first sample image and the second sample image into the key point detection model respectively to output the human key points in the first sample image and the second sample image; Setting weights for the human key points in the first sample image and the second sample image based on the difference between the human key points in the second sample image and the corresponding first sample image to obtain weighted key points; Fine-tuning the key point detection model based on the first sample image, the second sample image and the weighted key points.

5. The method according to claim 2, wherein The performing muscle fatigue analysis on the target object based on the initial key points of the image sequence and the personalized information of the target object to determine the fatigue key points includes: Inputting the initial key points of the image sequence and the personalized information of the target object into a muscle fatigue analysis model to output the fatigue key points of the target object, where the muscle fatigue analysis model is trained with a human posture sample image sequence, the personalized information of the human body and the muscle information generating fatigue.

6. The method according to claim 5, wherein The training steps of the muscle fatigue analysis model include: Inputting the human posture sample image sequence of a sample image sequence data set into a key point detection model to output the human key points in each image of the human posture sample image sequence; Analyzing the muscle electrical signals of the human body in the human posture sample image sequence or analyzing the actions of the human body in the human posture sample image sequence to determine the muscle information generating fatigue; Training the muscle fatigue analysis model based on the human key points of the human posture sample image sequence, the personalized information of the human body and the muscle information generating fatigue.

7. The method according to claim 1, wherein the posture prompt information includes key point position modification information and target action generation information, and the method further includes: Determining the posture change frequency of the target object based on the initial key points of the image sequence; Generating the posture prompt information based on the fatigue key points, including: Determining the posture type of the target object based on the posture change frequency, wherein the posture type includes a static posture and a dynamic posture; In response to the posture type of the target object being a dynamic posture, generating the target action generation information based on the fatigue key points; In response to the posture type of the target object being a static posture, generating the key point position modification information based on the fatigue key points.

8. The method according to claim 7, wherein Generating a target image for guiding the target object to adjust the posture based on the posture prompt information, including: Inputting the posture prompt information and a reference image into a second artificial intelligence content generation model, and outputting a target image for guiding the target object to adjust the posture.

9. The method according to claim 8, wherein Inputting the posture prompt information and a reference image into a second artificial intelligence content generation model, and outputting a target image for guiding the target object to adjust the posture, including: In response to the posture type of the target object being a static posture, inputting a frame image in the image sequence as the reference image into the encoder of the second artificial intelligence content generation model, and outputting a noisy image; Inputting the key point position modification information into the key point control module of the second artificial intelligence content generation model, and outputting key point position control information; Inputting the noisy image and the key point position control information into the decoder of the second artificial intelligence content generation model, and outputting a frame image for guiding the target object to adjust the posture.

10. The method according to claim 8, wherein, Inputting the posture prompt information and a reference image into a second artificial intelligence content generation model, and outputting a target image for guiding the target object to adjust the posture, including: In response to the posture type of the target object being a dynamic posture, obtaining a frame image from a reference image database as the reference image and inputting it into the encoder of the second artificial intelligence content generation model, and outputting a noisy image; Inputting the target action generation information into the time domain control module of the second artificial intelligence content generation model, and outputting action time domain control information; Inputting the noisy image and the action time domain control information into the decoder of the second artificial intelligence content generation model, and outputting a video for guiding the target object to adjust the posture.

11. A posture guidance image generation device, including: A fatigue detection module configured to determine the fatigue key points of the target object based on an image sequence including the target object; A posture adjustment module configured to generate posture prompt information based on the fatigue key points; An image generation module configured to generate a target image for guiding the target object to adjust the posture based on the posture prompt information.

12. An electronic device, characterized in that, Including: At least one processor; And A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1-10.

13. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are for causing the computer to perform the method according to any one of claims 1-10.

14. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1-10.

Images