Detection method for human body parts, action recognition method, device and electronic device

Abstract

The present application provides a method for detecting human body parts, a method for action recognition, a device and an electronic device. The method for detecting human body parts includes: obtaining a human body image to be detected; extracting feature information of the human body parts from the human body image; encoding the human body image based on the feature information to generate a target human body image; and detecting a part boundary box of the human body parts from the target human body image. Thus, the feature information of the human body parts can be extracted from the human body image, the human body image can be encoded based on the feature information of the human body parts to generate a target human body image, and the part boundary box of the human body parts can be detected from the target human body image, which helps to improve the accuracy and applicability of human body part detection.

Description

Technical Field

[0001] The present application relates to the field of computer application technologies, and particularly to a method for detecting human body parts, a method for recognizing actions, a device, an electronic device, and a storage medium. Background Art

[0002] Human body part detection refers to the technology of locating human body parts in an image. Currently, with the development of artificial intelligence technology, human body part detection has been widely applied. For example, in an intelligent monitoring scenario, human body part detection can be performed on an image collected by a camera, and whether a human behavior is abnormal can be identified based on the position of the human body part, so as to achieve intelligent monitoring and alarm of human behaviors. However, the human body part detection in the prior art has the problem of low accuracy. Summary of the Invention

[0003] The present application aims to at least partly solve one of the technical problems of low accuracy in the method for detecting human body parts in the related art.

[0004] To this end, an embodiment of the first aspect of the present application proposes a method for detecting human body parts, which can encode a human body image based on the feature information of the human body parts to generate a target human body image, and perform human body part detection based on the target human body image, which helps to improve the accuracy and applicability of human body part detection.

[0005] An embodiment of the second aspect of the present application proposes a method for recognizing actions, which can generate an action recognition result of any human body part, enrich the content of the action recognition result, and improve the fineness and efficiency of action recognition.

[0006] An embodiment of the third aspect of the present application proposes a device for detecting human body parts.

[0007] An embodiment of the fourth aspect of the present application proposes an action recognition device.

[0008] An embodiment of the fifth aspect of the present application proposes an electronic device.

[0009] An embodiment of the sixth aspect of the present application proposes a computer-readable storage medium.

[0010] An embodiment of the first aspect of the present application proposes a method for detecting human body parts, including: obtaining a human body image to be detected; extracting the feature information of the human body parts from the human body image; encoding the human body image based on the feature information to generate a target human body image; and detecting the part border of the human body parts from the target human body image.

[0011] According to the method for detecting human body parts according to an embodiment of the present application, feature information of human body parts can be extracted from a human body image, and the human body image can be encoded based on the feature information of the human body parts to generate a target human body image, and the part border of the human body parts can be detected from the target human body image. Thus, the human body image can be encoded based on the feature information of the human body parts to generate a target human body image, and human body part detection can be performed based on the target human body image, which helps to improve the accuracy and applicability of human body part detection.

[0012] In addition, the method for detecting human body parts according to the above embodiment of the present application may further have the following additional technical features:

[0013] In an embodiment of the present application, the feature information includes a target parameter for characterizing the area occupied by the human body part in the human body image, and the extracting the feature information from the human body image includes: obtaining the image area of the human body image and the ratio of the human body part to the human body; based on the image area and the ratio corresponding to any human body part, obtaining the target parameter of the any human body part.

[0014] In an embodiment of the present application, the human body parts include key points and / or skeletons, the skeleton is used to connect two key points, and the target parameter includes the radius of the key point and / or the width of the skeleton.

[0015] In an embodiment of the present application, the human body parts include key points and / or skeletons, the skeleton is used to connect two key points, and the target parameter includes the radius of the key point and / or the width of the skeleton.

[0016] In an embodiment of the present application, the feature information includes the category of the human body part, the position of the human body part in the human body image, and a target parameter for characterizing the area occupied by the human body part in the human body image; the determining the encoding color corresponding to the human body part and the encoding area in the human body image based on the feature information includes: determining the encoding color corresponding to the human body part based on the category of the human body part; determining the encoding area corresponding to the human body part based on the position and the target parameter of the human body part.

[0017] In an embodiment of the present application, the encoding the human body image based on the encoding color and the encoding area corresponding to each human body part includes: encoding the encoding area corresponding to any human body part based on the encoding color corresponding to the any human body part.

[0018] A second aspect embodiment of the present application proposes an action recognition method, including: obtaining a target recognition action of a human body part in a to-be-detected image; extracting feature information of the human body part from the to-be-detected image; encoding the to-be-detected image based on the feature information to generate a target image; detecting a part border of the human body part from the target image; associating the target recognition action of any human body part with the part border of the any human body part to generate an action recognition result of the any human body part.

[0019] According to the action recognition method of the embodiments of the present application, the feature information of the human body part can be extracted from the to-be-detected image, the to-be-detected image is encoded based on the feature information to generate a target image, the part border of the human body part is detected from the target image, and the target recognition action of any human body part is associated with the part border of the any human body part to generate an action recognition result. Thus, the to-be-detected image can be encoded based on the feature information of the human body part to generate a target image, and human body part detection is performed based on the target image, which helps to improve the accuracy and applicability of human body part detection, and the target recognition action of any human body part can be associated with the part border of the any human body part to generate an action recognition result of the any human body part, enriching the content of the action recognition result and improving the fineness of action recognition.

[0020] In addition, the action recognition method according to the above embodiments of the present application may further have the following additional technical features:

[0021] In an embodiment of the present application, the obtaining a target recognition action of a human body part in a to-be-detected image includes: obtaining a to-be-detected video and obtaining a target video segment from the to-be-detected video, where the target video segment includes the to-be-detected image; performing part action recognition on the target video segment to obtain the target recognition action of the human body part in the to-be-detected image.

[0022] In an embodiment of the present application, the encoding the to-be-detected image based on the feature information to generate a target image includes: based on the feature information, determining an encoding color corresponding to the human body part and an encoding area in the to-be-detected image; encoding the to-be-detected image based on the encoding color and the encoding area corresponding to each human body part to generate the target image.

[0023] In one embodiment of the present application, the feature information includes the category of the human body part, the position of the human body part in the image to be detected, and a target parameter for characterizing the area occupied by the human body part in the image to be detected; determining the encoding color corresponding to the human body part and the encoding area in the image to be detected based on the feature information includes: determining the encoding color corresponding to the human body part based on the category of the human body part; and determining the encoding area corresponding to the human body part based on the position and the target parameter of the human body part.

[0024] In one embodiment of the present application, the method further includes: obtaining a target recognition action of a human body and a human body frame in the image to be detected; and associating the target recognition action of any human body with the human body frame of the any human body to generate an action recognition result of the any human body.

[0025] In one embodiment of the present application, obtaining the target recognition action of the human body in the image to be detected includes: obtaining a video to be detected, where the video to be detected includes the image to be detected; and performing human body action recognition on the video to be detected to obtain the target recognition action of the human body in the image to be detected.

[0026] An embodiment of the third aspect of the present application provides a detection device for human body parts, including: an acquisition module, configured to acquire a human body image to be detected; an extraction module, configured to extract feature information of the human body part from the human body image; an encoding module, configured to encode the human body image based on the feature information to generate a target human body image; and a detection module, configured to detect a part frame of the human body part from the target human body image.

[0027] The detection device for human body parts according to the embodiment of the present application can extract the feature information of the human body part from the human body image, encode the human body image based on the feature information of the human body part to generate a target human body image, and detect the part frame of the human body part from the target human body image. Thus, the human body image can be encoded based on the feature information of the human body part to generate a target human body image, and human body part detection can be performed based on the target human body image, which helps to improve the accuracy and applicability of human body part detection.

[0028] In addition, the detection device for human body parts according to the above embodiments of the present application may further have the following additional technical features:

[0029] In an embodiment of the present application, the feature information includes a target parameter for characterizing the area occupied by the human body part in the human body image. The extraction module includes: a first acquisition unit for acquiring the image area of the human body image and the ratio of the human body part to the human body; a second acquisition unit for acquiring the target parameter of any human body part based on the image area and the ratio corresponding to any human body part.

[0030] In an embodiment of the present application, the human body part includes key points and / or a skeleton. The skeleton is used to connect two key points, and the target parameter includes the radius of the key point and / or the width of the skeleton.

[0031] In an embodiment of the present application, the encoding module includes: a determination unit for determining the encoding color corresponding to the human body part and the encoding area in the human body image based on the feature information; an encoding unit for encoding the human body image based on the encoding color and the encoding area corresponding to each human body part to generate the target human body image.

[0032] In an embodiment of the present application, the feature information includes the category of the human body part, the position of the human body part in the human body image, and a target parameter for characterizing the area occupied by the human body part in the human body image. The determination unit is further configured to: determine the encoding color corresponding to the human body part based on the category of the human body part; determine the encoding area corresponding to the human body part based on the position and the target parameter of the human body part.

[0033] In an embodiment of the present application, the encoding unit is further configured to: encode the encoding area corresponding to any human body part based on the encoding color corresponding to any human body part.

[0034] An embodiment of the fourth aspect of the present application provides an action recognition device, including: a first acquisition module for acquiring a target recognition action of a human body part in a to-be-detected image; an extraction module for extracting feature information of the human body part from the to-be-detected image; an encoding module for encoding the to-be-detected image based on the feature information to generate a target image; a detection module for detecting a part border of the human body part from the target image; a first association module for associating the target recognition action of any human body part with the part border of the any human body part to generate an action recognition result of the any human body part.

[0035] The action recognition device according to the embodiment of the present application can extract the feature information of the human body part from the image to be detected, encode the image to be detected based on the feature information to generate a target image, detect the part border of the human body part from the target image, and associate the target recognition action of any human body part with the part border of any human body part to generate an action recognition result. Thus, the image to be detected can be encoded based on the feature information of the human body part to generate a target image, and human body part detection can be performed based on the target image, which helps to improve the accuracy and applicability of human body part detection, and the target recognition action of any human body part can be associated with the part border of any human body part to generate the action recognition result of any human body part, enriching the content of the action recognition result and improving the fineness of action recognition.

[0036] In addition, the action recognition device according to the above embodiment of the present application may further have the following additional technical features:

[0037] In an embodiment of the present application, the first acquisition module includes: a first acquisition unit for acquiring a video to be detected and acquiring a target video segment from the video to be detected, where the target video segment includes the image to be detected; a second acquisition unit for performing part action recognition on the target video segment to acquire the target recognition action of the human body part in the image to be detected.

[0038] In an embodiment of the present application, the encoding module includes: a determination unit for determining the encoding color corresponding to the human body part and the encoding region in the image to be detected based on the feature information; an encoding unit for encoding the image to be detected based on the encoding color and the encoding region corresponding to each human body part to generate the target image.

[0039] In an embodiment of the present application, the feature information includes the category of the human body part, the position of the human body part in the image to be detected, and a target parameter for characterizing the area occupied by the human body part in the image to be detected; the determination unit is further configured to: determine the encoding color corresponding to the human body part based on the category of the human body part; and determine the encoding region corresponding to the human body part based on the position and the target parameter of the human body part.

[0040] In an embodiment of the present application, the device further includes: a second acquisition module for acquiring the target recognition action of the human body and the human body border in the image to be detected; a second association module for associating the target recognition action of any human body with the human body border of any human body to generate the action recognition result of any human body.

[0041] In one embodiment of the present application, the second acquisition module includes: a third acquisition unit configured to acquire a video to be detected, where the video to be detected includes the image to be detected; and an identification unit configured to perform human action recognition on the video to be detected to obtain a target recognition action of the human body in the image to be detected.

[0042] An embodiment of the fifth aspect of the present application provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the method for detecting human body parts as described in the embodiment of the first aspect above, or implements the action recognition method as described in the embodiment of the second aspect above.

[0043] The electronic device according to the embodiment of the present application can extract feature information of human body parts from a human body image through the processor executing a computer program stored on the memory, and encode the human body image based on the feature information of the human body parts to generate a target human body image, and can detect the part border of the human body parts from the target human body image. Thus, the human body image can be encoded based on the feature information of the human body parts to generate a target human body image, and human body part detection can be performed based on the target human body image, which helps to improve the accuracy and applicability of human body part detection.

[0044] An embodiment of the sixth aspect of the present application provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for detecting human body parts as described in the embodiment of the first aspect above, or implements the action recognition method as described in the embodiment of the second aspect above.

[0045] The computer-readable storage medium according to the embodiment of the present application can extract feature information of human body parts from a human body image by storing a computer program and executing it by a processor, and encode the human body image based on the feature information of the human body parts to generate a target human body image, and can detect the part border of the human body parts from the target human body image. Thus, the human body image can be encoded based on the feature information of the human body parts to generate a target human body image, and human body part detection can be performed based on the target human body image, which helps to improve the accuracy and applicability of human body part detection.

[0046] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present invention. Description of the Drawings

[0047] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of the embodiments in conjunction with the drawings, where:

[0048] Figure 1Schematic flowchart of a method for detecting human body parts according to an embodiment of the present application;

[0049] Figure 2 Schematic diagrams of a human body detector and a human body key part detector based on pose guidance;

[0050] Figure 3 Schematic flowchart of extracting feature information of human body parts from a human body image in a method for detecting human body parts according to an embodiment of the present application;

[0051] Figure 4 Schematic flowchart of encoding a human body image based on feature information to generate a target human body image in a method for detecting human body parts according to an embodiment of the present application;

[0052] Figure 5 Schematic flowchart of an action recognition method according to an embodiment of the present application;

[0053] Figure 6 Schematic diagrams of video action prediction and key part action prediction;

[0054] Figure 7 Schematic diagram of obtaining a target recognition action of a human body part in a human body image in an action recognition method according to an embodiment of the present application;

[0055] Figure 8 Schematic structural diagram of a device for detecting human body parts according to an embodiment of the present application;

[0056] Figure 9 Schematic structural diagram of an action recognition device according to an embodiment of the present application; and

[0057] Figure 10 Schematic structural diagram of an electronic device according to an embodiment of the present application. Detailed implementation manners

[0058] The embodiments of the present application will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions from beginning to end. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present application, but should not be construed as limiting the present application.

[0059] The methods, devices, electronic devices, and storage media for detecting human body parts and action recognition according to the embodiments of the present application will be described below with reference to the accompanying drawings.

[0060] Figure 1 Schematic flowchart of a method for detecting human body parts according to an embodiment of the present application.

[0061] AsFigure 1 As shown in the figure, the method for detecting human body parts according to the embodiment of the present application includes:

[0062] S101. Obtain a human body image to be detected.

[0063] It should be noted that the execution subject of the method for detecting human body parts according to the embodiment of the present application can be a detection device for human body parts. The detection device for human body parts according to the embodiment of the present application can be configured in any electronic device so that the electronic device can execute the method for detecting human body parts according to the embodiment of the present application. Among them, the electronic device can be a personal computer (PC for short), a cloud device, a mobile device, etc. The mobile device can be, for example, a mobile phone, a tablet computer, a personal digital assistant, a wearable device, a vehicle-mounted device, etc., which are hardware devices with various operating systems, touch screens, and / or display screens.

[0064] In the embodiment of the present application, the human body image may include at least one human body. For example, a human body image may include one human body, or a human body image may include multiple human bodies.

[0065] In one implementation, obtaining a human body image to be detected may include obtaining an image to be detected, detecting a human body border from the image to be detected, and using the image to be detected within the human body border as the human body image to be detected.

[0066] In one implementation, as Figure 2 shown in the figure, detecting a human body border from the image to be detected may include inputting the image to be detected X f into a trained human body detector, and the human body detector detects a human body border (not shown in the figure) from the image to be detected X f . Further, the image to be detected X f within the human body border can be used as the human body image X p to be detected. It should be noted that the type of the human body detector is not overly limited. For example, the human body detector includes but is not limited to R-CNN (Region Convolutional Neural Networks), YOLOF (You Only Look One-level Feature), etc.

[0067] S102. Extract feature information of human body parts from the human body image.

[0068] It should be noted that the type of human body parts is not overly limited. For example, human body parts include but are not limited to the head, left hand, right hand, left arm, right arm, buttocks, left foot, right foot, left leg, right leg, etc.

[0069] It should be noted that the category of the feature information of the human body part is not overly limited. For example, the feature information of the human body part includes, but is not limited to, the category of the human body part, the position of the human body part in the human body image, the target parameter for characterizing the area occupied by the human body part in the human body image, etc.

[0070] In one implementation, the human body part includes at least one of key points and skeletons, and the skeleton is used to connect two key points. It should be noted that the category of the key points is not overly limited. For example, the key points can include joint key points, skeleton key points, etc. It can be understood that a human body part can include at least one key point, and / or a human body part can include at least one skeleton.

[0071] In one implementation, the position of the human body part in the human body image can include the position set of the key points and / or skeletons included in the human body part in the human body image, and the target parameter includes the radius of the key point and / or the width of the skeleton.

[0072] In one implementation, as Figure 2 shown, extracting the feature information of the human body part from the human body image can include inputting the human body image X p into the trained pose estimator, and the pose estimator extracts the feature information K of the human body part from the human body image. p It should be noted that the category of the pose estimator is not overly limited. For example, the pose estimator can include HRNet (High-Resoultion Net).

[0073] S103, Encode the human body image based on the feature information to generate the target human body image.

[0074] In one implementation, encoding the human body image based on the feature information to generate the target human body image can include encoding the human body image based on the feature information of the key points and / or skeletons included in the human body part to generate the target human body image.

[0075] In one implementation, encoding the human body image based on the feature information to generate the target human body image can include generating RGB dots based on the feature information of any human body part, and encoding the human body image based on the RGB dots to generate the target human body image. It can be understood that different feature information can correspond to different RGB dots. Among them, generating RGB dots based on the feature information of any human body part can include obtaining the color and radius of the RGB dots based on the feature information of any human body part, and generating RGB dots based on the color and radius of the RGB dots.

[0076] In one implementation, as Figure 2As shown, encoding the human body image based on the feature information to generate a target human body image may include using the feature information K p and the human body image X p as inputs to a pose encoder, and the pose encoder encodes the human body image X p based on the feature information K p to generate the target human body image X t .

[0077] S104. Detect the part bounding boxes of the human body parts from the target human body image.

[0078] It can be understood that the number of part bounding boxes of the human body parts detected from the target human body image can be multiple, and one human body part can correspond to one part bounding box. It should be noted that no excessive limitation is imposed on the shape of the part bounding box of the human body part. For example, the part bounding box includes but is not limited to rectangles, squares, circles, etc. There may be overlapping areas between the part bounding boxes of different human body parts.

[0079] In one embodiment, as Figure 2 shown, detecting the part bounding boxes of the human body parts from the target human body image may include inputting the target human body image X t into a trained human body part detector, and the human body part detector detects the part bounding boxes of the human body parts from the target human body image X t . As Figure 2 shown, the part bounding boxes of the head, left hand, right hand, hip, left leg, and right leg can be detected from the target human body image X t . It should be noted that no excessive limitation is imposed on the type of the human body part detector. For example, the human body part detector includes but is not limited to R-CNN (Region Convolutional Neural Networks), YOLOF (You Only Look One-level Feature), etc.

[0080] In summary, according to the method for detecting human body parts in the embodiments of the present application, the feature information of the human body parts can be extracted from the human body image, and the human body image is encoded based on the feature information of the human body parts to generate a target human body image, and the part bounding boxes of the human body parts are detected from the target human body image. Thus, the human body image can be encoded based on the feature information of the human body parts to generate a target human body image, and human body part detection is performed based on the target human body image, which helps to improve the accuracy and applicability of human body part detection.

[0081] Based on any of the above embodiments, as Figure 3 shown, extracting the feature information of the human body parts from the human body image in step S102 includes:

[0082] S301. Obtain the image area of the human body image and the ratio of the human body part to the human body.

[0083] In one implementation, obtaining the image area of the human body image may include obtaining the product of the width and length of the human body image, and using the above product as the image area of the human body image. When the width of the human body image is the pixel width and the length of the human body image is the pixel length, the obtained image area of the human body image is the pixel area.

[0084] It should be noted that the ratio of the human body part to the human body refers to the ratio of a human body part in the whole human body. It can be understood that different human body parts in one human body may correspond to different ratios, and the same human body parts in different human bodies may correspond to different ratios.

[0085] In one implementation, obtaining the ratio of the human body part to the human body may include identifying the category of the human body and obtaining the ratio of the human body part to the human body according to the category of the human body. It should be noted that there are no excessive limitations on the category of the human body. For example, the category of the human body includes but is not limited to male, female, child, young adult, elderly, etc. It can be understood that for different categories of human bodies, the corresponding ratios of human body parts to the human body are also different.

[0086] In one implementation, obtaining the ratio of the human body part to the human body according to the category of the human body may include pre - establishing a mapping relationship or mapping table between the category of the human body, the human body part and the ratio of the human body part to the human body. After obtaining the category of the human body, querying the above mapping relationship or mapping table can obtain the ratio of the human body part mapped to the category of the human body. It should be noted that there are no excessive limitations on the above mapping relationship or mapping table.

[0087] S302. Based on the image area and the ratio corresponding to any human body part, obtain the target parameter of any human body part.

[0088] In one implementation, based on the image area and the ratio corresponding to any human body part, obtaining the target parameter of any human body part may include obtaining the radius of the key points included in any human body part and / or the width of the skeleton based on the image area and the ratio corresponding to the key points and / or the skeleton included in any human body part.

[0089] In one implementation, based on the image area and the ratio corresponding to any human body part, obtaining the target parameter of any human body part may include obtaining the product of the image area and the ratio corresponding to any human body part, and determining the target parameter of any human body part based on the above product.

[0090] For example, when the human body part is the right hand, the right hand includes key points A, B, C and skeletons 1, 2. Skeleton 1 is used to connect key points A and B, and skeleton 2 is used to connect key points B and C. The radius of key point A can be determined based on the product of the image area and the ratio corresponding to key point A. The radius of key point B can be determined based on the product of the image area and the ratio corresponding to key point B. The radius of key point C can be determined based on the product of the image area and the ratio corresponding to key point C. The width of skeleton 1 can be determined based on the product of the image area and the ratio corresponding to skeleton 1. The width of skeleton 2 can be determined based on the product of the image area and the ratio corresponding to skeleton 2.

[0091] Thus, based on the image area and the ratio corresponding to any human body part, the method can obtain the target parameters of any human body part.

[0092] Based on any of the above embodiments, as Figure 4 shown, encoding the human body image based on the feature information in step S103 to generate a target human body image includes:

[0093] S401, based on the feature information, determine the encoding color corresponding to the human body part and the encoding area in the human body image.

[0094] It can be understood that different feature information may correspond to different encoding colors and encoding areas.

[0095] In one implementation, based on the feature information, determining the encoding color corresponding to the human body part may include determining the encoding color corresponding to the human body part based on the category of the human body part. It can be understood that different categories of human body parts may correspond to different encoding colors. For example, the encoding colors corresponding to the head, right hand, and left hand are red, yellow, and green respectively.

[0096] In one implementation, based on the feature information, determining the encoding color corresponding to the human body part may include determining the encoding color corresponding to the key points and / or skeletons based on the category of the key points and / or skeletons included in the human body part. For example, when the human body part is the right hand, the right hand includes key points A, B, C and skeletons 1, 2. Skeleton 1 is used to connect key points A and B, and skeleton 2 is used to connect key points B and C. The encoding colors corresponding to key points A, B, C can be determined respectively based on the categories of key points A, B, C, and the encoding colors corresponding to skeletons 1, 2 can be determined respectively based on the categories of skeletons 1, 2.

[0097] It should be noted that the encoding areas corresponding to different human body parts do not overlap.

[0098] In one implementation, based on the feature information, determining the encoding area in the human body image may include determining the encoding area corresponding to the human body part based on the position and target parameters of the human body part.

[0099] In one embodiment, to determine the encoding region corresponding to a human body part based on the position of the human body part and the target parameter may include determining the area occupied by the human body part in the human body image based on the target parameter, and taking the region that spreads outward from the position of the human body part in the human body image according to the above area as the encoding region.

[0100] In one embodiment, to determine the encoding region in the human body image based on the feature information may include determining the encoding regions corresponding to the key points and / or the skeleton based on the positions of the key points and / or the skeleton included in the human body part and the target parameter. It should be noted that the encoding region corresponding to the human body part includes the encoding regions corresponding to the key points and / or the skeleton.

[0101] For example, when the human body part is the right hand, the right hand includes key points A, B, C and skeletons 1, 2. Skeleton 1 is used to connect key points A and B, and skeleton 2 is used to connect key points B and C. The areas occupied by key points A, B, C in the human body image can be determined respectively based on the radii of key points A, B, C, and the areas occupied by skeletons 1, 2 in the human body image can be determined respectively based on the widths of skeletons 1, 2. Take the region that spreads outward from the position of key point A in the human body image according to the area corresponding to key point A as the encoding region of key point A. For the content of determining the encoding regions of key points B, C, and skeletons 1, 2, reference can be made to the content of determining the encoding region of key point A above, which will not be elaborated here. It should be noted that the encoding region of the right hand may include the encoding regions corresponding to key points A, B, C, and skeletons 1, 2.

[0102] S402. Encode the human body image based on the encoding color and encoding region corresponding to each human body part to generate a target human body image. In one embodiment, to encode the human body image based on the encoding color and encoding region corresponding to each human body part may include encoding the encoding region corresponding to any human body part based on the encoding color corresponding to any human body part.

[0103] For example, if the encoding colors corresponding to the head, right hand, and left hand are red, yellow, and green respectively, and the encoding regions corresponding to the head, right hand, and left hand are encoding regions 1, 2, and 3 respectively, then encoding region 1 can be encoded based on red, encoding region 2 can be encoded based on yellow, and encoding region 3 can be encoded based on green.

[0104] In one embodiment, to encode the human body image based on the encoding color and encoding region corresponding to each human body part may include encoding the human body image based on the encoding colors and encoding regions corresponding to the key points and / or the skeleton included in each human body part.

[0105] For example, when the human body part is the right hand, the right hand includes key points A, B, C and skeletons 1, 2. Skeleton 1 is used to connect key points A and B, and skeleton 2 is used to connect key points B and C. The encoding colors corresponding to key points A, B, and C are red, yellow, and green respectively, and the encoding colors corresponding to skeletons 1 and 2 are both blue. The encoding regions corresponding to key points A, B, C, skeletons 1, and 2 are encoding regions 1 to 5 respectively. Then, encoding region 1 can be encoded based on red, encoding region 2 can be encoded based on yellow, encoding region 3 can be encoded based on green, and encoding regions 4 and 5 can be encoded based on blue.

[0106] Thus, in this method, the encoding color corresponding to the human body part and the encoding regions in the human body image can be determined based on the feature information, and the human body image can be encoded based on the encoding color and encoding regions corresponding to each human body part to generate the target human body image.

[0107] Figure 5 It is a schematic flow chart of an action recognition method according to an embodiment of the present application.

[0108] As Figure 5 shown, the action recognition method of the embodiment of the present application includes:

[0109] S501, obtaining the target recognition action of the human body part in the image to be detected.

[0110] It should be noted that the execution subject of the action recognition method of the embodiment of the present application can be an action recognition device. The action recognition device of the embodiment of the present application can be configured in any electronic device so that the electronic device can execute the action recognition method of the embodiment of the present application. Among them, the electronic device can be a personal computer (Personal Computer, abbreviated as PC), a cloud device, a mobile device, etc. The mobile device can be, for example, a mobile phone, a tablet computer, a personal digital assistant, a wearable device, a vehicle-mounted device, etc., which are hardware devices with various operating systems, touch screens, and / or display screens.

[0111] It can be understood that the target recognition actions of different human body parts may be different.

[0112] In one implementation, obtaining the target recognition action of the human body part in the image to be detected may include performing part action recognition on the image to be detected to obtain the target recognition action of the human body part in the image to be detected.

[0113] In one implementation, obtaining the target recognition action of the human body part in the image to be detected may include obtaining the image to be detected video and obtaining the target video segment from the image to be detected video. Among them, the target video segment includes the image to be detected, and performing part action recognition on the target video segment to obtain the target recognition action of the human body part in the image to be detected.

[0114] In one embodiment, as Figure 6 shown, performing part action recognition on a target video segment may include inputting the target video segment into a trained part action recognition model, and the part action recognition model performs part action recognition on the target video segment. It should be noted that the category of the part action recognition model is not overly limited.

[0115] In one embodiment, performing part action recognition on a target video segment may include selecting at least one first candidate frame image from multiple first frame images included in the target video segment, performing part action recognition on any one of the first candidate frame images, obtaining candidate recognition actions of human body parts in any one of the first candidate frame images, and determining target recognition actions of human body parts in the image to be detected based on the candidate recognition actions of human body parts in each first candidate frame image.

[0116] In one embodiment, the selection method of at least one first candidate frame image is random selection.

[0117] In one embodiment, determining target recognition actions of human body parts in the image to be detected based on the candidate recognition actions of human body parts in each first candidate frame image may include, for any human body part, determining the candidate recognition action with the most occurrences as the target recognition action. For example, if the candidate recognition actions of the right hand in the first candidate frame images A, B, and C are grip, release, and grip respectively, then grip can be determined as the target recognition action of the right hand.

[0118] In one embodiment, target recognition actions of human body parts in each frame image of the video to be detected can be determined based on the candidate recognition actions of human body parts in each first candidate frame image. Thus, it is not necessary to perform part action recognition on each frame image in the video to be detected in this method, which helps to save computing resources.

[0119] S502, extracting feature information of human body parts from the image to be detected.

[0120] S503, encoding the image to be detected based on the feature information to generate a target image.

[0121] In one embodiment, encoding the image to be detected based on the feature information to generate a target image may include, based on the feature information, determining the encoding color corresponding to the human body part and the encoding region in the image to be detected, and encoding the image to be detected based on the encoding color and encoding region corresponding to each human body part to generate a target image.

[0122] In one embodiment, the feature information includes the category of the human body part, the position of the human body part in the image to be detected, and a target parameter for characterizing the area occupied by the human body part in the image to be detected. Based on the feature information, the encoded color corresponding to the human body part and the encoded area in the image to be detected are determined, which may include determining the encoded color corresponding to the human body part based on the category of the human body part, and determining the encoded area corresponding to the human body part based on the position and the target parameter of the human body part.

[0123] S504. Detect the part border of the human body part from the target image.

[0124] It should be noted that the relevant content of steps S502 - S504 can be referred to the above embodiments and will not be elaborated here.

[0125] S505. Associate the target recognition action of any human body part with the part border of any human body part to generate the action recognition result of any human body part.

[0126] For example, the target recognition actions of the head, left hand, and right hand are target recognition actions A, B, and C respectively, and the part borders of the head, left hand, and right hand are part borders 1 to 3 respectively. The target recognition action A can be associated with part border 1 to generate the action recognition result of the head, the target recognition action B can be associated with part border 2 to generate the action recognition result of the left hand, and the target recognition action C can be associated with part border 3 to generate the action recognition result of the right hand.

[0127] In one embodiment, associating the target recognition action of any human body part with the part border of any human body part may include associating the first identifier of the target recognition action of any human body part with the second identifier of the part border of any human body part. It should be noted that no excessive limitations are imposed on the categories of the first identifier and the second identifier. For example, the first identifier and the second identifier include but are not limited to names, numbers, characters, etc.

[0128] In one embodiment, associating the target recognition action of any human body part with the part border of any human body part may include constructing at least one of the corresponding relationship, mapping relationship, and mapping table between the target recognition action of any human body part and the part border of any human body part.

[0129] In summary, according to the action recognition method of the embodiments of the present application, the feature information of the human body parts can be extracted from the image to be detected, the image to be detected is encoded based on the feature information to generate a target image, the part border of the human body part is detected from the target image, and the target recognition action of any human body part is associated with the part border of any human body part to generate an action recognition result. Thus, the image to be detected can be encoded based on the feature information of the human body part to generate a target image, and the human body part detection is performed based on the target image, which helps to improve the accuracy and applicability of the human body part detection, and the target recognition action of any human body part can be associated with the part border of any human body part to generate the action recognition result of any human body part, enriching the content of the action recognition result and improving the fineness of the action recognition.

[0130] Based on any of the above embodiments, as Figure 7 shown, obtaining the action recognition result of the image to be detected may include:

[0131] S701, obtaining the target recognition action of the human body and the human body border in the image to be detected;

[0132] It should be noted that for the relevant content of obtaining the human body border in the image to be detected, reference can be made to the above embodiments, and details are not described herein again.

[0133] In one implementation manner, obtaining the target recognition action of the human body in the image to be detected may include performing human body action recognition on the image to be detected to obtain the target recognition action of the human body in the image to be detected.

[0134] In one implementation manner, obtaining the target recognition action of the human body in the image to be detected may include obtaining the video to be detected, where the video to be detected includes the image to be detected, and performing human body action recognition on the video to be detected to obtain the target recognition action of the human body in the image to be detected.

[0135] In one implementation manner, as Figure 6 shown, performing human body action recognition on the video to be detected may include inputting the video to be detected into a trained human body action recognition model, and the human body action recognition model performs human body action recognition on the video to be detected. It should be noted that the category of the human body action recognition model is not overly limited.

[0136] In one implementation manner, performing human body action recognition on the video to be detected may include selecting at least one second candidate frame image from the multiple second frame images included in the video to be detected, performing human body action recognition on any second candidate frame image to obtain the candidate recognition action of the human body in any second candidate frame image, and determining the target recognition action of the human body in the image to be detected based on the candidate recognition action of the human body in each second candidate frame image.

[0137] In one implementation, the target recognition action of the human body in each frame image of the video to be detected can be determined based on the candidate recognition actions of the human body in each second candidate frame image. For example, the video action label of the video to be detected can be determined based on the candidate recognition actions of the human body in each second candidate frame image. Continuing with Figure 6 as an example, the video action label of the video to be detected is a throwing motion. Thus, in this method, it is not necessary to perform human body action recognition on each frame image in the video to be detected, which helps to save computing resources.

[0138] S702. Associate the target recognition action of any human body with the human body bounding box of any human body to generate the action recognition result of any human body.

[0139] For example, if the target recognition actions of human bodies 1 to 3 are target recognition actions A, B, and C respectively, and the human body bounding boxes of human bodies 1 to 3 are human body bounding boxes 1 to 3 respectively, the target recognition action A can be associated with the human body bounding box 1 to generate the action recognition result of human body 1, the target recognition action B can be associated with the human body bounding box 2 to generate the action recognition result of human body 2, and the target recognition action C can be associated with the human body bounding box 3 to generate the action recognition result of human body 3.

[0140] It should be noted that for the relevant content of associating the target recognition action of any human body with the human body bounding box of any human body, reference can be made to the above embodiments, which will not be elaborated here.

[0141] Thus, in this method, the target recognition action of any human body can be associated with the human body bounding box of any human body to generate the action recognition result of any human body, enriching the content of the action recognition result and improving the fineness of action recognition.

[0142] Corresponding to the human body part detection method provided in the above Figures 1 to 4 embodiment, the present disclosure also provides a human body part detection device. Since the human body part detection device provided in the embodiments of the present disclosure corresponds to the human body part detection method provided in the above Figures 1 to 4 embodiment, the implementation manners of the human body part detection method are also applicable to the human body part detection device provided in the embodiments of the present disclosure and will not be described in detail in the embodiments of the present disclosure.

[0143] Figure 8 FIG. is a structural schematic diagram of a human body part detection device according to an embodiment of the present application.

[0144] As Figure 8 shown, the human body part detection device 100 of the embodiments of the present application may include: an acquisition module 110, an extraction module 120, an encoding module 130, and a detection module 140.

[0145] An acquisition module 110, configured to acquire a human body image to be detected.

[0146] An extraction module 120, configured to extract feature information of a human body part from the human body image.

[0147] An encoding module 130, configured to encode the human body image based on the feature information to generate a target human body image.

[0148] A detection module 140, configured to detect a part border of the human body part from the target human body image.

[0149] In an embodiment of the present application, the feature information includes a target parameter for characterizing the area occupied by the human body part in the human body image. The extraction module 120 includes: a first acquisition unit, configured to acquire the image area of the human body image and the ratio of the human body part to the human body; a second acquisition unit, configured to acquire the target parameter of any human body part based on the image area and the ratio corresponding to any human body part.

[0150] In an embodiment of the present application, the human body part includes key points and / or a skeleton, the skeleton is used to connect two key points, and the target parameter includes the radius of the key point and / or the width of the skeleton.

[0151] In an embodiment of the present application, the encoding module 130 includes: a determination unit, configured to determine the encoding color corresponding to the human body part and the encoding area in the human body image based on the feature information; an encoding unit, configured to encode the human body image based on the encoding color and the encoding area corresponding to each human body part to generate the target human body image.

[0152] In an embodiment of the present application, the feature information includes the category of the human body part, the position of the human body part in the human body image, and a target parameter for characterizing the area occupied by the human body part in the human body image. The determination unit is further configured to: determine the encoding color corresponding to the human body part based on the category of the human body part; determine the encoding area corresponding to the human body part based on the position and the target parameter of the human body part.

[0153] In an embodiment of the present application, the encoding unit is further configured to: encode the encoding area corresponding to any human body part based on the encoding color corresponding to any human body part.

[0154] The human body part detection device according to the embodiment of the present application can extract the feature information of the human body part from the human body image, encode the human body image based on the feature information of the human body part to generate a target human body image, and detect the part border of the human body part from the target human body image. Thus, the human body image can be encoded based on the feature information of the human body part to generate a target human body image, and the human body part detection can be performed based on the target human body image, which helps to improve the accuracy and applicability of the human body part detection.

[0155] Corresponding to the Figures 5 to 7 action recognition method provided in the above Figures 5 to 7 embodiment, the present disclosure also provides an action recognition device. Since the action recognition device provided in the embodiment of the present disclosure corresponds to the

[0156] Figure 9 action recognition method provided in the above

[0157] As Figure 9 shown, the action recognition device 200 according to the embodiment of the present application may include: a first acquisition module 210, an extraction module 220, an encoding module 230, a detection module 240, and a first association module 250.

[0158] The first acquisition module 210 is configured to acquire the target recognition action of the human body part in the image to be detected.

[0159] The extraction module 220 is configured to extract the feature information of the human body part from the image to be detected.

[0160] The encoding module 230 is configured to encode the image to be detected based on the feature information to generate a target image.

[0161] The detection module 240 is configured to detect the part border of the human body part from the target image.

[0162] The first association module 250 is configured to associate the target recognition action of any human body part with the part border of the any human body part to generate the action recognition result of the any human body part.

[0163] In an embodiment of the present application, the first acquisition module 210 includes: a first acquisition unit configured to acquire an image to be detected video and acquire a target video segment from the image to be detected video, where the target video segment includes the image to be detected; a second acquisition unit configured to perform part action recognition on the target video segment to acquire the target recognition action of the human body part in the image to be detected.

[0164] In one embodiment of the present application, the encoding module 230 includes: a determination unit, configured to determine the encoding color corresponding to the human body part and the encoding region in the image to be detected based on the feature information; an encoding unit, configured to encode the image to be detected based on the encoding color and the encoding region corresponding to each human body part, and generate the target image.

[0165] In one embodiment of the present application, the feature information includes the category of the human body part, the position of the human body part in the image to be detected, and a target parameter for characterizing the area occupied by the human body part in the image to be detected; the determination unit is further configured to: determine the encoding color corresponding to the human body part based on the category of the human body part; and determine the encoding region corresponding to the human body part based on the position and the target parameter of the human body part.

[0166] In one embodiment of the present application, the apparatus further includes: a second acquisition module, configured to acquire the target recognition action and the human body frame of the human body in the image to be detected; a second association module, configured to associate the target recognition action of any human body with the human body frame of the any human body, and generate the action recognition result of the any human body.

[0167] In one embodiment of the present application, the second acquisition module includes: a third acquisition unit, configured to acquire a video to be detected, where the video to be detected includes the image to be detected; and an identification unit, configured to perform human body action recognition on the video to be detected to acquire the target recognition action of the human body in the image to be detected.

[0168] The action recognition apparatus according to the embodiments of the present application can extract the feature information of the human body part from the image to be detected, encode the image to be detected based on the feature information to generate the target image, detect the part frame of the human body part from the target image, associate the target recognition action of any human body part with the part frame of any human body part, and generate the action recognition result. Thus, the image to be detected can be encoded based on the feature information of the human body part to generate the target image, and the human body part can be detected based on the target image, which helps to improve the accuracy and applicability of the human body part detection, and the target recognition action of any human body part can be associated with the part frame of any human body part to generate the action recognition result of any human body part, enriching the content of the action recognition result and improving the fineness of the action recognition.

[0169] To implement the above embodiments, as Figure 10As shown in the figure, the present application also provides an electronic device 300, including: a memory 310, a processor 320, and a computer program stored on the memory 310 and executable on the processor 320. When the processor 320 executes the program, it implements the method for detecting human body parts proposed in the foregoing embodiments of the present application, or implements the method for action recognition proposed in the foregoing embodiments of the present application.

[0170] In the electronic device according to the embodiment of the present application, by executing the computer program stored in the memory through the processor, the feature information of the human body parts can be extracted from the human body image, and the image can be encoded based on the feature information to generate a target human body image, and the part border of the human body parts can be detected from the target human body image. Thus, the part border of the human body parts can be detected, realizing the prediction of the actions of more fine-grained human key parts and improving the fineness and efficiency of human body part detection.

[0171] To implement the above embodiments, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for detecting human body parts proposed in the foregoing embodiments of the present application, or implements the method for action recognition proposed in the foregoing embodiments of the present application.

[0172] In the computer-readable storage medium according to the embodiment of the present application, by storing the computer program and executing it by the processor, the feature information of the human body parts can be extracted from the human body image, and the image can be encoded based on the feature information to generate a target human body image, and the part border of the human body parts can be detected from the target human body image. Thus, the part border of the human body parts can be detected, realizing the prediction of the actions of more fine-grained human key parts and improving the fineness and efficiency of human body part detection.

[0173] In the description of this specification, the descriptions with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, without conflict, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0174] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In the description of the present application, the meaning of "a plurality" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0175] Any process or method description represented in a flowchart or otherwise described herein can be understood to represent a module, segment, or portion of code including one or more executable instructions for implementing a customized logic function or process. The scope of the preferred embodiments of the present application includes additional implementations where functions may be executed in a substantially simultaneous manner or in an order opposite to that shown or discussed, according to the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application pertain.

[0176] The logic and / or steps represented in a flowchart or otherwise described herein, for example, can be considered a sequenced list of executable instructions for implementing a logical function, which can be specifically implemented in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of the computer-readable medium include the following: an electrical connection portion with one or more wirings (electronic device), a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, as the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or other appropriate processing as necessary, and then stored in a computer memory.

[0177] It should be understood that each part of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one of the following techniques known in the art or a combination thereof can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0178] Those of ordinary skill in the art can understand that all or part of the steps carried by the method of implementing the above embodiments can be completed by instructing relevant hardware through a program. The said program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0179] In addition, in each embodiment of the present application, each functional unit can be integrated into a processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The above integrated module can be implemented in the form of hardware or in the form of a software functional module. When the above integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0180] The above-mentioned storage medium can be a read-only memory, a magnetic disk, an optical disc, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A method for detecting human body parts, characterized in that, it includes: Obtain a human body image to be detected; Extract the feature information of the human body parts from the human body image; Encode the human body image based on the feature information to generate a target human body image; Detect the part border of the human body part from the target human body image; The encoding the human body image based on the feature information to generate a target human body image includes: Based on the feature information, determine the encoding color corresponding to the human body part and the encoding area in the human body image, wherein different feature information corresponds to different encoding colors and encoding areas; Encode the human body image based on the encoding color and the encoding area corresponding to each human body part to generate the target human body image.

2. The method according to claim 1, characterized in that, The feature information includes a target parameter for characterizing the area occupied by the human body part in the human body image, and the extracting the feature information from the human body image includes: Obtain the image area of the human body image and the ratio of the human body part to the human body; Based on the image area and the ratio corresponding to any human body part, obtain the target parameter of the any human body part.

3. The method according to claim 2, characterized in that, The human body parts include key points and / or skeletons, the skeleton is used to connect two key points, and the target parameters include the radius of the key points and / or the width of the skeleton.

4. The method according to claim 1, characterized in that, The feature information includes the category of the human body part, the position of the human body part in the human body image, and a target parameter for characterizing the area occupied by the human body part in the human body image; The determining the encoding color corresponding to the human body part and the encoding area in the human body image based on the feature information includes: Based on the category of the human body part, determine the encoding color corresponding to the human body part; Based on the position and the target parameter of the human body part, determine the encoding area corresponding to the human body part.

5. The method according to claim 1, characterized in that, The encoding the human body image based on the encoding color and the encoding area corresponding to each human body part includes: Encode the encoding area corresponding to any human body part based on the encoding color corresponding to the any human body part.

6. An action recognition method, characterized in that, it includes: Obtain the target recognition action of the human body part in the image to be detected; Extract the feature information of the human body part from the image to be detected; Encode the image to be detected based on the feature information to generate a target image; Detect the part border of the human body part from the target image; Associate the target recognition action of any human body part with the part border of the any human body part to generate the action recognition result of the any human body part; The encoding the image to be detected based on the feature information to generate a target image includes: Based on the feature information, determine the encoded color corresponding to the human body part and the encoded area in the image to be detected, where different feature information corresponds to different encoded colors and encoded areas; Based on the encoded color and the encoded area corresponding to each human body part, encode the image to be detected to generate the target image.

7. The method according to claim 6, wherein, the obtaining the target recognition action of the human body part in the image to be detected includes: obtain the video to be detected, and obtain the target video segment from the video to be detected, where the target video segment includes the image to be detected; perform part action recognition on the target video segment to obtain the target recognition action of the human body part in the image to be detected.

8. The method according to claim 6, wherein, the feature information includes the category of the human body part, the position of the human body part in the image to be detected, and a target parameter for characterizing the area occupied by the human body part in the image to be detected; the determining the encoded color corresponding to the human body part and the encoded area in the image to be detected based on the feature information includes: determine the encoded color corresponding to the human body part based on the category of the human body part; determine the encoded area corresponding to the human body part based on the position and the target parameter of the human body part.

9. The method according to claim 6, wherein, the method further includes: obtain the target recognition action of the human body and the human body border in the image to be detected; associate the target recognition action of any human body with the human body border of the any human body to generate the action recognition result of the any human body.

10. The method according to claim 9, wherein, the obtaining the target recognition action of the human body in the image to be detected includes: obtain the video to be detected, where the video to be detected includes the image to be detected; perform human action recognition on the video to be detected to obtain the target recognition action of the human body in the image to be detected.

11. A detection device for a human body part, wherein, comprises: an obtaining module, configured to obtain a human body image to be detected; an extraction module, configured to extract the feature information of the human body part from the human body image; an encoding module, configured to encode the human body image based on the feature information to generate a target human body image; a detection module, configured to detect the part border of the human body part from the target human body image; the encoding module includes: a determination unit, configured to determine the encoded color corresponding to the human body part and the encoded area in the human body image based on the feature information, where different feature information corresponds to different encoded colors and encoded areas; an encoding unit, configured to encode the human body image based on the encoded color and the encoded area corresponding to each human body part to generate the target human body image.

12. The device according to claim 11, wherein, The feature information includes a target parameter for characterizing the area occupied by the human body part in the human body image, and the extraction module includes: A first acquisition unit, configured to acquire the image area of the human body image and the ratio of the human body part to the human body; A second acquisition unit, configured to acquire the target parameter of any human body part based on the image area and the ratio corresponding to any human body part.

13. The apparatus according to claim 12, wherein, the human body part includes key points and / or a skeleton, the skeleton is used to connect two key points, and the target parameter includes the radius of the key point and / or the width of the skeleton.

14. The apparatus according to claim 11, wherein, the feature information includes the category of the human body part, the position of the human body part in the human body image, and a target parameter for characterizing the area occupied by the human body part in the human body image; the determination unit is further configured to: determine the encoded color corresponding to the human body part based on the category of the human body part; determine the encoded area corresponding to the human body part based on the position and the target parameter of the human body part.

15. The apparatus according to claim 11, wherein, the encoding unit is further configured to: encode the encoded area corresponding to any human body part based on the encoded color corresponding to any human body part.

16. An action recognition apparatus, wherein, it includes: A first acquisition module, configured to acquire a target recognition action of a human body part in a to-be-detected image; An extraction module, configured to extract feature information of the human body part from the to-be-detected image; An encoding module, configured to encode the to-be-detected image based on the feature information to generate a target image; A detection module, configured to detect a part border of the human body part from the target image; A first association module, configured to associate the target recognition action of any human body part with the part border of the any human body part to generate an action recognition result of the any human body part; the encoding module includes: A determination unit, configured to determine an encoded color corresponding to the human body part and an encoded area in the to-be-detected image based on the feature information, wherein different feature information corresponds to different encoded colors and encoded areas; An encoding unit, configured to encode the to-be-detected image based on the encoded color and the encoded area corresponding to each human body part to generate the target image.

17. The apparatus according to claim 16, wherein, the first acquisition module includes: A first acquisition unit, configured to acquire a to-be-detected video and acquire a target video segment from the to-be-detected video, wherein the target video segment includes the to-be-detected image; A second acquisition unit, configured to perform part action recognition on the target video segment to acquire the target recognition action of the human body part in the to-be-detected image.

18. The apparatus according to claim 16, wherein, The feature information includes the category of the human body part, the position of the human body part in the image to be detected, and a target parameter for characterizing the area occupied by the human body part in the image to be detected; The determining unit is further configured to: determine the encoded color corresponding to the human body part based on the category of the human body part; determine the encoded area corresponding to the human body part based on the position and the target parameter of the human body part.

19. The apparatus according to claim 16, wherein, the apparatus further comprises: a second obtaining module, configured to obtain a target recognition action and a human body frame of a human body in the image to be detected; a second association module, configured to associate the target recognition action of any human body with the human body frame of the any human body to generate an action recognition result of the any human body.

20. The apparatus according to claim 19, wherein, the second obtaining module comprises: a third obtaining unit, configured to obtain a video to be detected, where the video to be detected includes the image to be detected; a recognition unit, configured to perform human body action recognition on the video to be detected to obtain the target recognition action of the human body in the image to be detected.

21. An electronic device, wherein, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, when the processor executes the program, implementing the method for detecting a human body part according to any one of claims 1-5, or implementing the method for action recognition according to any one of claims 6-10.

22. A computer-readable storage medium, on which a computer program is stored, wherein, when the program is executed by a processor, implementing the method for detecting a human body part according to any one of claims 1-5, or implementing the method for action recognition according to any one of claims 6-10.

Images