Image processing, training, visual task processing method, device, equipment and medium

Abstract

The present disclosure provides an image processing, training, visual task processing method, device, equipment and medium, relates to the technical field of image processing, and particularly relates to the technical field of image annotation, industrial quality inspection, computer vision and deep learning. The specific implementation scheme is: in response to detecting an annotation operation for a first target image, displaying a first annotation image in a first annotation interface, the first annotation image including an annotation region, first display size information of the annotation region being determined according to predetermined ratio information and first actual size information of the annotation region; determining annotation path information corresponding to the first annotation image according to the first annotation image, the annotation path information including position information of at least one annotation pixel; and sending the predetermined ratio information and the annotation path information to a server, so that the server obtains image recovery information corresponding to the first annotation image according to the predetermined ratio information and the annotation path information.

Description

Technical Field

[0001] This disclosure relates to the field of image processing technology, and more particularly to the fields of image annotation, industrial quality inspection, computer vision, and deep learning technology. Specifically, it relates to an image processing, training, and visual task processing method, apparatus, device, and medium. Background Technology

[0002] With the development of computer technology, artificial intelligence technology has also advanced. Examples include deep learning and computer vision technologies. To ensure the usability of AI models, labeled images can be used to train them. Summary of the Invention

[0003] This disclosure provides an image processing, training, and visual task processing method, apparatus, device, and medium.

[0004] According to one aspect of this disclosure, an image processing method is provided, comprising: in response to detecting an annotation operation on a first target image, displaying a first annotation image on a first annotation interface, wherein the first annotation image includes an annotation region, the annotation region includes at least one annotation pixel, first display size information of the annotation region is determined based on predetermined ratio information and first actual size information of the annotation region, the predetermined ratio information including a predetermined ratio greater than 0 and less than 1; determining annotation path information corresponding to the first annotation image based on the first annotation image, wherein the annotation path information includes position information of the at least one annotation pixel; and sending the predetermined ratio information and the annotation path information to a server, so that the server obtains image restoration information corresponding to the first annotation image based on the predetermined ratio information and the annotation path information.

[0005] According to another aspect of this disclosure, an image processing method is provided, comprising: in response to receiving predetermined ratio information and annotation path information from a client, generating image restoration information corresponding to a first annotated image based on the predetermined ratio information and the annotation path information; wherein the annotation path information is determined by the client based on the predetermined ratio information and the first annotated image, the first annotated image is displayed on a first annotation interface by the client in response to detecting an annotation operation on a first target image, the first annotated image includes an annotation region, the annotation region includes at least one annotation pixel, and the annotation path information includes position information of a portion of the annotation pixels; wherein the first display size information of the annotation region is determined based on the predetermined ratio information and the first actual size information of the annotation region, and the predetermined ratio information includes a predetermined ratio greater than 0 and less than 1.

[0006] According to another aspect of the present disclosure, a training method of a visual task processing model is provided, comprising: obtaining a display image; and training a deep learning model using the display image to obtain the visual task processing model; wherein the display image comprises one of a first display image and a second display image, the first display image comprises a second labeled image and a second target image, the second display image comprises a third labeled image and a third target image, the first display image is obtained according to the method described above, and the second display image is obtained according to the method described above.

[0007] According to another aspect of the present disclosure, a visual task processing method is provided, comprising: obtaining a visual image to be processed; and processing the visual image to be processed using a visual task processing model to obtain visual processing information; wherein the visual task processing model is trained using the method described above.

[0008] According to another aspect of the present disclosure, an image processing device is provided, comprising: a first display module configured to display a first labeled image in a first labeling interface in response to detecting a labeling operation on a first target image, wherein the first labeled image comprises a labeling region, the labeling region comprises at least one labeled pixel, and first display size information of the labeling region is determined according to predetermined ratio information and first actual size information of the labeling region, and the predetermined ratio information comprises a predetermined ratio greater than 0 and less than 1; a first determination module configured to determine labeling path information corresponding to the first labeled image according to the first labeled image, wherein the labeling path information comprises position information of the at least one labeled pixel; and a first sending module configured to send the predetermined ratio information and the labeling path information to a server, so that the server obtains image recovery information corresponding to the first labeled image according to the predetermined ratio information and the labeling path information.

[0009] According to another aspect of the present disclosure, an image processing apparatus is provided, comprising: a second determining module configured to, in response to receiving predetermined ratio information and annotation path information from a client, determine image restoration information corresponding to a first annotation image according to the predetermined ratio information and the annotation path information; wherein the annotation path information is determined by the client according to the predetermined ratio information and the first annotation image, the first annotation image is displayed by the client in response to detecting an annotation operation on a first target image in a first annotation interface, the first annotation image comprises an annotation region, the annotation region comprises at least one annotation pixel, and the annotation path information comprises position information of the at least one annotation pixel; wherein first display size information of the annotation region is determined according to the predetermined ratio information and first actual size information of the annotation region, and the predetermined ratio information comprises a predetermined ratio greater than 0 and less than 1.

[0010] According to another aspect of the present disclosure, a training apparatus of a visual task processing model is provided, comprising: a first obtaining module configured to obtain a display image; and a first obtaining module configured to train a deep learning model using the display image to obtain the visual task processing model; wherein the display image comprises one of a first display image and a second display image, the first display image comprises a second annotation image and a second target image, the second display image comprises a third annotation image and a third target image, the first display image is obtained according to the apparatus of the present disclosure, and the second display image is obtained according to the apparatus of the present disclosure.

[0011] According to another aspect of the present disclosure, a visual task processing apparatus is provided, comprising: a second obtaining module configured to obtain a visual image to be processed; and a second obtaining module configured to process the visual image to be processed using a visual task processing model to obtain visual processing information; wherein the visual task processing model is trained using the apparatus of the present disclosure.

[0012] According to another aspect of the present disclosure, an electronic device is provided, comprising: a plurality of processors; and a memory in communication with the plurality of processors; wherein the memory stores instructions executable by the plurality of processors, and the instructions are executed by the plurality of processors to enable the plurality of processors to perform the method as described in the present disclosure.

[0013] According to another aspect of the present disclosure, a non-transitory computer readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to perform the method as described in the present disclosure.

[0014] According to another aspect of the present disclosure, there is provided a computer program product comprising a computer program which, when executed by a processor, implements the method as described above in the present disclosure.

[0015] It should be understood that the matters described herein are intended to be illustrative rather than limiting. For example, while the application is illustrated and described in relation to a computer system, the application is not limited to computer systems. The application can be implemented in any suitable environment. Such can be appreciated by one skilled in the art and the same is contemplated herein. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings are included to provide a further understanding of the application, and are incorporated in and constitute a part of this specification. In the drawings:

[0017] Figure 1 An exemplary system architecture to which the image processing method, the training method of the visual task processing model, the visual task processing method and the device according to embodiments of the present disclosure can be applied is schematically shown;

[0018] Figure 2 A flowchart of the image processing method according to embodiments of the present disclosure is schematically shown;

[0019] Figure 3 An example schematic diagram showing that a first annotation image is displayed on a first annotation interface in response to detecting an annotation operation for a first target image according to embodiments of the present disclosure is schematically shown;

[0020] Figure 4 An example schematic diagram showing that at least one pixel to be traversed in the first annotation image is determined according to predetermined ratio information according to embodiments of the present disclosure is schematically shown;

[0021] Figure 5 A flowchart of the image processing method according to another embodiment of the present disclosure is schematically shown;

[0022] Figure 6 A signaling diagram of the image processing method according to embodiments of the present disclosure is schematically shown;

[0023] Figure 7 A flowchart of the training method of the visual task processing model according to embodiments of the present disclosure is schematically shown;

[0024] Figure 8 A flowchart of the visual task processing method according to embodiments of the present disclosure is schematically shown;

[0025] Figure 9 An example schematic diagram of the visual task processing method according to embodiments of the present disclosure is schematically shown;

[0026] Figure 10 A block diagram of the image processing device according to embodiments of the present disclosure is schematically shown;

[0027] Figure 11 a block diagram of an image processing apparatus according to another embodiment of the present disclosure is shown schematically;

[0028] Figure 12 a block diagram of a training apparatus of a visual task processing model according to an embodiment of the present disclosure is shown schematically;

[0029] Figure 13 a block diagram of a visual task processing apparatus according to an embodiment of the present disclosure is shown schematically; and

[0030] Figure 14 a block diagram of an electronic device suitable for implementing the image processing method, the training method of the visual task processing model and the visual task processing method according to an embodiment of the present disclosure is shown schematically. DETAILED DESCRIPTION

[0031] Exemplary embodiments of the present disclosure are described herein with reference to the accompanying drawings, in which various details of embodiments of the present disclosure are set forth to help the understanding of the present disclosure. It should be understood, however, that the embodiments described herein are merely exemplary and are not intended to limit the scope of the present disclosure. It will be apparent to those of ordinary skill in the art that various changes and modifications can be made thereto without departing from the scope and spirit of the present disclosure. Also, for the sake of brevity and clarity, descriptions of well-known functions and constructions are omitted herein.

[0032] Figure 1 An exemplary system architecture to which the image processing method, the training method of the visual task processing model, the visual task processing method and apparatus according to an embodiment of the present disclosure can be applied is shown schematically.

[0033] It should be noted that Figure 1 The system architecture shown is merely an example of a system architecture to which embodiments of the present disclosure can be applied, to help those skilled in the art understand the technical content of the present disclosure, but does not mean that embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios.

[0034] As Figure 1 shown, the system architecture 100 according to this embodiment can include a first client 101, a second client 102, a third client 103, a network 104 and a server 105. The network 104 is a medium for providing a communication link between the first client 101, the second client 102, the third client 103 and the server 105. The network 104 can include various connection types. For example, at least one of wired and wireless communication links. The client can include at least one of the first client 101, the second client 102 and the third client 103.

[0035] A user can use at least one of the first client 101, the second client 102, and the third client 103 to interact with the server 105 through the network 104 to receive or send messages, etc. At least one of the first client 101, the second client 102, and the third client 103 can be installed with various communication client applications. For example, at least one of a knowledge reading type application, a web browser application, a search type application, an instant messaging tool, an email client, and a social platform software, etc.

[0036] The first client 101, the second client 102, and the third client 103 can be various electronic devices with a display screen and supporting web browsing. For example, the electronic devices can include at least one of a smart phone, a tablet computer, a laptop computer, and a desktop computer, etc.

[0037] The server 105 can be a server providing various services. For example, the server 105 can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system to solve the defects of large management difficulty and weak business scalability in traditional physical hosts and VPS services (Virtual Private Server).

[0038] For example, the first client 101 can display a first labeled image on the first labeling interface in response to detecting a labeling operation on the first target image. The first labeled image includes a labeling region. The labeling region can include at least one labeled pixel. The first display size information of the labeling region can be determined according to the predetermined ratio information and the first actual size information of the labeling region. The predetermined ratio information includes a predetermined ratio, which can be greater than 0 and less than 1. According to the predetermined ratio information and the first labeled image, labeling path information corresponding to the first labeled image is determined. The labeling path information can include position information of at least one labeled pixel. The above can be executed by a web browser of the first client 101.

[0039] The server 105 can determine image recovery information corresponding to the first labeled image according to the predetermined ratio information and the labeling path information in response to receiving the predetermined ratio information and the labeling path information from the first client 101.

[0040] It should be noted that the visual task processing method provided in the embodiments of the present disclosure can be executed by one of the first terminal device 101, the second terminal device 102, and the third terminal device 103. Correspondingly, the visual task processing apparatus provided in the embodiments of the present disclosure can also be arranged in one of the first terminal device 101, the second terminal device 102, and the third terminal device 103.

[0041] Alternatively, the visual task processing method provided by the embodiments of the present disclosure can also be generally executed by the server 105. Correspondingly, the visual task processing apparatus provided by the embodiments of the present disclosure can be generally arranged in the server 105. The visual task processing method provided by the embodiments of the present disclosure can also be executed by a server or a server cluster different from the server 105 and capable of communicating with at least one of the first terminal device 101, the second terminal device 102, the third terminal device 103 and the server 105. Correspondingly, the visual task processing apparatus provided by the embodiments of the present disclosure can also be arranged in a server or a server cluster different from the server 105 and capable of communicating with at least one of the first terminal device 101, the second terminal device 102, the third terminal device 103 and the server 105.

[0042] It should be noted that the training method of the visual task processing model provided by the embodiments of the present disclosure can be generally executed by the server 105. Correspondingly, the training apparatus of the visual task processing model provided by the embodiments of the present disclosure can be generally arranged in the server 105. The training method of the visual task processing model provided by the embodiments of the present disclosure can also be executed by a server or a server cluster different from the server 105 and capable of communicating with at least one of the first terminal device 101, the second terminal device 102, the third terminal device 103 and the server 105. Correspondingly, the training apparatus of the visual task processing model provided by the embodiments of the present disclosure can also be arranged in a server or a server cluster different from the server 105 and capable of communicating with at least one of the first terminal device 101, the second terminal device 102, the third terminal device 103 and the server 105.

[0043] Alternatively, the visual task processing method provided by the embodiments of the present disclosure can also be generally executed by one of the first terminal device 101, the second terminal device 102 and the third terminal device 103. Correspondingly, the visual task processing apparatus provided by the embodiments of the present disclosure can also be arranged in one of the first terminal device 101, the second terminal device 102 and the third terminal device 103.

[0044] It should be understood that Figure 1 The number of the first client, the second client, the third client, the network and the server in the above method is only illustrative. According to the implementation needs, there can be any number of the first client, the second client, the third client, the network and the server.

[0045] It should be noted that the serial numbers of the various operations in the following method are only used to represent the operations for description, and should not be regarded as representing the execution sequence of the various operations. Unless explicitly indicated, the method does not need to be executed in the order shown.

[0046] Figure 2A flowchart of an image processing method according to an embodiment of the present disclosure is schematically shown.

[0047] As shown in Figure 2 The method 200 includes operations S210-S230.

[0048] At operation S210, in response to detecting a labeling operation on the first target image, a first labeled image is displayed on a first labeling interface.

[0049] At operation S220, according to the first labeled image, labeling path information corresponding to the first labeled image is determined.

[0050] At operation S230, the predetermined ratio information and the labeling path information are sent to a server, so that the server obtains image recovery information corresponding to the first labeled image according to the predetermined ratio information and the labeling path information.

[0051] According to an embodiment of the present disclosure, the first labeled image can include a labeling region. The labeling region can include at least one labeled pixel. The first display size information of the labeling region can be determined according to the predetermined ratio information and the first actual size information of the labeling region. The predetermined ratio included in the predetermined ratio information can be greater than 0 and less than 1. The labeling path information includes position information of the at least one labeled pixel.

[0052] According to an embodiment of the present disclosure, the labeling operation can be used to generate labeling data corresponding to the labeling region. The labeling operation can include one of a regular labeling operation and an irregular labeling operation. The regular labeling operation can refer to a labeling operation with regular shape. For example, the regular labeling operation can include at least one of a point labeling operation, a polyline labeling operation, and a polygon labeling operation. The irregular labeling operation can refer to a labeling operation with irregular shape. For example, the irregular labeling operation can include a smear labeling operation. The smear labeling operation can be a labeling operation with a smear diameter. The labeling operation of the embodiment of the present disclosure can include the smear labeling operation.

[0053] According to an embodiment of the present disclosure, the annotation interface can refer to an interface for providing an annotation operation. The annotation interface can refer to a viewport of a browser. The annotation interface can be layered. For example, the annotation interface can include an annotation layer and an image layer. The annotation interface can be a canvas. The canvas can include Canvas. The annotation data can be stored in the form of an annotated image. The annotated image can include at least one of a non-annotation region and at least one annotation region. The annotation region can refer to a region through which an annotation operation passes. The number of annotation regions can be configured according to actual business needs, which is not limited herein. For example, the annotated image can include 10 annotation regions. The 10 annotation regions can overlap each other or can not overlap each other. The non-annotation region can refer to a region through which the annotation operation does not pass. The annotation region can include at least one annotation pixel. The annotation pixel can refer to a pixel that is annotated. The non-annotation region can include at least one non-annotation pixel. The non-annotation pixel can refer to a pixel that is not annotated. The pixel value of the annotation pixel and the pixel value of the non-annotation pixel can be different. For example, the pixel value of the annotation pixel is a non-zero value. The pixel value of the non-annotation pixel is a zero value. The annotation region can be generated by an operation body performing an annotation operation. The operation body can include at least one of a mouse, a stylus, and the like. In the case where the annotation interface is a first annotation interface, the first annotation interface can have the same functions and structures as the annotation interface, which is not described herein again.

[0054] According to an embodiment of the present disclosure, the size information can include a width (i.e., Width) and a height (i.e., Height). The size information of the image can include at least one of actual size information of the image and display size information of the image. The actual size information of the image can refer to actual size information of the image. The display size information of the image can refer to size information of the image displayed. The actual size information of the annotation interface can refer to actual size information of the annotation interface. The display size information of the annotation region can refer to size information of the annotation region displayed. The actual size information of the annotation region can refer to actual size information of the annotation region. The display size information can be referred to as current size information. The first annotation interface can be used to display the first annotated image. In addition, the first annotation interface can also be used to display the first target image.

[0055] According to an embodiment of the present disclosure, the first target image can refer to an object on which an annotation operation is performed. The first annotated image can be an image generated in response to an annotation operation on the first target image. The first annotated image can include at least one annotation region. In addition, the first annotated image can also include a non-annotation region. The annotation region can include at least one annotation pixel. The non-annotation region can include at least one non-annotation pixel. The first target image can serve as a background image of the first annotated image.

[0056] According to an embodiment of the present disclosure, the first display size information can refer to display size information of the labeling region. The first actual size information can refer to actual size information of the labeling region. The fifth display size information can refer to display size information of the first target image. The fifth actual size information can refer to actual size information of the first target image. The first display size information can include a first width and a first height. The first actual size information can include a second width and a second height. The fifth display size information can include a fifth width and a fifth height. The fifth actual size information can include a sixth width and a sixth height.

[0057] According to an embodiment of the present disclosure, the first target image can be an image obtained by adjusting the second actual size information of the second target image. For example, the first target image can be an image obtained by reducing the second actual size information of the second target image, so that the fifth display size information is smaller than the second actual size information, and the fifth actual size information is smaller than the second actual size information. The fifth display size information and the fifth actual size information can be target size information. The target size information can be size information matching the third actual size information. For example, the target size information can be size information consistent with the third actual size information, that is, the third actual size information can be determined as the target size information. The tenth display size information and the tenth actual size information of the first labeling image can be the same, for example, the tenth display size information and the tenth actual size information of the first labeling image are both matched with the third actual size information.

[0058] According to an embodiment of the present disclosure, the predetermined ratio information can include a first predetermined ratio and a second predetermined ratio. The first predetermined ratio and the second predetermined ratio can be the same or different. The first predetermined ratio can be a number greater than 0 and less than 1. The second predetermined ratio can be a number greater than 0 and less than 1. The first predetermined ratio can refer to predetermined ratio information corresponding to the width. The second predetermined ratio can refer to predetermined ratio information corresponding to the height. The first display size information can be determined according to the predetermined ratio information and the first actual size information. For example, the first display size information can be determined according to the product between the predetermined ratio information and the first actual size information. The first width can be determined according to the product between the first predetermined ratio and the second width. The first height can be determined according to the product between the second predetermined ratio and the second height.

[0059] According to an embodiment of the present disclosure, the format of the first labeling image can be configured according to actual business needs, which is not limited here and can be downloaded and displayed by the client relatively quickly. For example, the format of the first labeling image can include Base64. Since the color of the first labeling image in the Base64 format is single and the volume is small, it can be downloaded and displayed by the client relatively quickly.

[0060] According to an embodiment of the present disclosure, the annotation path can refer to a path formed by the annotation region. The annotation path can include at least one target annotation pixel. The at least one target annotation pixel can have a time sequence relationship. The annotation path information can be used to describe the position information of the annotation path. The annotation path information can refer to the position information corresponding to the annotation path. The annotation path information can be determined according to the position information of the at least one target annotation pixel. The at least one target annotation pixel can be at least part of the at least one annotation pixel included in the annotation region. The position information of the target annotation pixel satisfies a predetermined position condition. The pixel value of the target annotation pixel satisfies a predetermined pixel value condition. The predetermined position condition can be used to determine whether the position information is the position information of the to-be-traversed pixel. The predetermined pixel value condition can be used to determine whether the pixel value of the to-be-traversed pixel is a non-predetermined pixel value. The to-be-traversed pixel can refer to a pixel that needs to be traversed.

[0061] According to an embodiment of the present disclosure, the image restoration information can be used to generate an annotation image corresponding to the first annotation image. For example, a second annotation image corresponding to the first annotation image can be generated according to the image restoration information. The sixth actual size information and the sixth display size information of the second annotation image are the same. The sixth actual size information and the sixth display size information can both be the same as the second actual size information of the second target image.

[0062] According to an embodiment of the present disclosure, the client can detect an annotation operation on the first target image, and in a case where the annotation operation on the first target image is detected, an annotation region corresponding to the annotation operation can be obtained. According to the annotation region corresponding to the annotation operation, a first annotation image can be generated. The first annotation image can be displayed on the first annotation interface.

[0063] According to an embodiment of the present disclosure, the client can generate annotation path information according to the first annotation image. For example, the client can traverse the first annotation image by using a compression traversal method to obtain annotation path information corresponding to the first annotation image. A compression multiple corresponding to the compression traversal method is determined according to predetermined ratio information. The compression multiple represents a ratio between a first to-be-traversed pixel number and a total pixel number. The first to-be-traversed pixel number represents a sum of the number of the first to-be-traversed pixels in the first annotation image. The total pixel number represents a sum of the number of the pixels in the first annotation image. Alternatively, the client can determine at least one second to-be-traversed pixel in the first annotation image. According to the position information of the at least one second to-be-traversed pixel, the annotation path information corresponding to the first annotation image can be obtained. The second to-be-traversed pixel number is less than or equal to the total pixel number. The second to-be-traversed pixel number represents a sum of the number of the at least one second to-be-traversed pixel. For a second to-be-traversed pixel in the at least one second to-be-traversed pixel, there can be a first to-be-traversed pixel consistent with the second to-be-traversed pixel in the at least one first to-be-traversed pixel.

[0064] According to an embodiment of the present disclosure, after determining the annotation path information, the client can send the predetermined ratio information and the annotation path information to the server. The server can generate the image restoration information corresponding to the first annotation image according to the predetermined ratio information and the annotation path information. The image restoration information can be used to generate the second annotation image corresponding to the first annotation image.

[0065] According to an embodiment of the present disclosure, the image processing method described above can be executed by a client. The client can be one of the first client 101, the second client 102 and the third client 103 in the system 100. Figure 1

[0066] According to an embodiment of the present disclosure, since the first actual size information of the annotation region included in the first annotation image is determined according to the predetermined ratio information and the second actual size information of the second target image corresponding to the first target image, and the predetermined ratio information includes a predetermined ratio greater than 0 and less than 1, the annotation region of the first target image is reduced, and the annotation efficiency for the first target image is further improved. In addition, since the annotation path information corresponding to the first annotation image is determined according to the first annotation image, the amount of calculation for determining the annotation path information can be reduced, and the storage speed of the annotation path information is improved. Since the server generates the image restoration information corresponding to the first annotation image according to the predetermined ratio information and the annotation path information without the participation of the client, the memory space of the client such as a browser is released, and the problem of client such as browser lag is alleviated.

[0067] According to an embodiment of the present disclosure, the predetermined ratio information can be determined according to the second actual size information of the second target image corresponding to the first target image and the third actual size information of the first annotation interface.

[0068] According to an embodiment of the present disclosure, the second actual size information can be actual size information of the second target image. The third actual size information can refer to actual size information of the first annotation interface. The fifth display size information can refer to display size information of the first target image. The fifth actual size information can refer to actual size information of the first target image. The second actual size information can include a third width and a third height. The third actual size information can include a fourth width and a fourth height. The second actual size information can satisfy a predetermined size condition. For example, the third width included in the second actual size information is greater than or equal to a predetermined width threshold. The third height included in the second actual size information is greater than or equal to a predetermined height threshold. The predetermined width threshold and the predetermined height threshold can be configured according to actual business requirements, which are not limited herein.

[0069] ​According to an embodiment of the present disclosure, the predetermined ratio information can include a first predetermined ratio and a second predetermined ratio. The second actual size information can include a third width and a third height. The third actual size information can include a fourth width and a fourth height. The third width is greater than the fourth width. The third height is greater than the fourth height. The first predetermined ratio can be determined according to a ratio between the third width and the fourth width. The second predetermined ratio can be determined according to a ratio between the third height and the fourth height. The first predetermined ratio and the second predetermined ratio can be the same or different. The first predetermined ratio and the second predetermined ratio are greater than 0 and less than 1.

[0070] According to an embodiment of the present disclosure, since the first actual size information of the annotation region included in the first annotation image is determined according to the predetermined ratio information and the second actual size information of the second target image corresponding to the first target image, and the predetermined ratio information includes a predetermined ratio greater than 0 and less than 1, the first annotation image can be adapted to the third actual size information of the first annotation interface by reducing the second actual size information of the second target image, thereby reducing the annotation region of the first target image, and further improving the annotation efficiency for the first target image. In addition, since the annotation path information corresponding to the first annotation image is determined according to the first annotation image, the actual size information of the first annotation image is reduced relative to the second actual size information of the second target image, so that the amount of calculation for determining the annotation path information can be reduced, and the storage speed of the annotation path information is improved. Since the server generates the image recovery information corresponding to the first annotation image according to the predetermined ratio information and the annotation path information without the participation of the client, the memory space of the client such as a browser is released, and the problem of client such as browser lag is alleviated.

[0071] According to an embodiment of the present disclosure, the image processing method described above can further include the following operations.

[0072] The second actual size information of the second target image is adjusted to the third actual size information to obtain the first target image.

[0073] According to an embodiment of the present disclosure, the second actual size information can be greater than the third actual size information. The second actual size information can be reduced to the third actual size information to obtain the first target image. The fifth display size information and the fifth actual size information of the first target image can both be consistent with the third actual size information.

[0074] According to an embodiment of the present disclosure, since the first target image is obtained by adjusting the second actual size information of the second target image to the third actual size information, the actual size information of the first target image is the third actual size information. Since the third actual size information is smaller than the second actual size information of the second target image, the first annotation image obtained by performing the annotation operation on the first target image can be adapted to the third actual size information of the first annotation interface by reducing the second actual size information of the second target image, thereby reducing the annotation region of the first target image and further improving the annotation efficiency for the first target image.

[0075] According to an embodiment of the present disclosure, operation S210 can include the following operations.

[0076] In response to detecting a mouse start event for the first target image, it is determined that the annotation operation for the first target image is triggered. In response to detecting a mouse end event for the first target image, it is determined that the annotation operation for the first target image is ended. In response to detecting that the annotation operation for the first target image is ended, a first annotation image is generated according to an annotation region generated in a process from triggering the annotation operation to ending the annotation operation. The first annotation image is displayed on an annotation layer.

[0077] According to an embodiment of the present disclosure, the annotation layer can be located in the first annotation interface.

[0078] According to an embodiment of the present disclosure, the mouse start event can refer to an event for triggering the mouse to perform the annotation operation. The mouse end event can refer to an event for ending the mouse to perform the annotation operation. Triggering the mouse to perform the annotation operation can refer to triggering the annotation operation. Ending the mouse to perform the annotation operation can refer to ending the annotation operation. For example, in a case where it is detected that the mouse starts to act on the first annotation interface, it is determined that the mouse start event is triggered. In a case where it is detected that the mouse does not act on the first annotation interface, it is determined that the mouse end event is triggered.

[0079] According to an embodiment of the present disclosure, the client can generate the first annotation image according to an annotation region formed by the mouse in a process from triggering the annotation operation to ending the annotation operation in a case where it is detected that the annotation operation for the first target image is ended. The first annotation image can be displayed on an annotation layer of the annotation interface.

[0080] According to an embodiment of the present disclosure, by detecting the mouse start event and the mouse end event to determine the triggering and ending of the annotation operation, and generating the first annotation image according to an annotation region formed by the mouse in a process from triggering the annotation operation to ending the annotation operation, the first annotation image is automatically generated, and the generation efficiency of the first annotation image is improved.

[0081] According to embodiments of this disclosure, the second display size information of the annotation layer can be matched with the third actual size information.

[0082] According to embodiments of this disclosure, the second display size information of the annotation layer may be consistent with the third actual size information. For example, the second display size information of the annotation layer is the third actual size information.

[0083] According to embodiments of this disclosure, the first target image may be located in the image layer. The image layer may be located in the first annotation interface. The annotation layer may be located above the image layer. The third display size information of the image layer may be matched with the third actual size information.

[0084] According to embodiments of this disclosure, the third display size information of the image layer may be consistent with the third actual size information. For example, the third display size information of the image layer is the third actual size information.

[0085] According to embodiments of this disclosure, by setting a first target image on the image layer and a first annotation image corresponding to the first target image on the annotation layer, layered storage of image data and annotation data is achieved, which facilitates editing operations on the annotation data.

[0086] Figure 3 The illustration shows an example diagram of a first annotation interface displaying a first annotation image in response to the detection of an annotation operation for a first target image, according to an embodiment of the present disclosure.

[0087] like Figure 3 As shown, in 300, the second actual size information of the second target image 301 is 8000×8000. The third actual size information of the first annotation interface 302_1 of the client 302 is 400×400. The first annotation interface 302_1 may include an annotation layer 302_1_1 and an image layer 302_1_2.

[0088] The second actual size information of the second target image 301 is adjusted to the third actual size information of the first annotation interface 302_1 to obtain the first target image 303. In response to detecting an annotation operation on the first target image 303, the first annotation image 304 is displayed on the annotation layer 302_1_1, and the first target image 303 is displayed on the image layer 302_1_2. A third display image 305 can be obtained based on the first annotation image 304 and the first target image 303. The "solid black area" in the first annotation image 304 represents the annotation area.

[0089] According to the second actual size information and the third actual size information, the first predetermined ratio included in the predetermined ratio information is 20, and the second predetermined ratio is 20. Since the first predetermined ratio and the second predetermined ratio are 20, the time consumption in the labeling stage is shortened by 20 times, thereby improving the labeling efficiency.

[0090] According to an embodiment of the present disclosure, operation S220 can include the following operations.

[0091] At least one pixel to be traversed in the first labeled image is determined. Labeling path information corresponding to the first labeled image is obtained according to position information of the at least one pixel to be traversed.

[0092] According to an embodiment of the present disclosure, the number of pixels to be traversed can be less than or equal to the total number of pixels. The number of pixels to be traversed can represent the sum of the number of at least one pixel to be traversed. The total number of pixels can represent the sum of the number of pixels in the first labeled image.

[0093] According to an embodiment of the present disclosure, the pixel to be traversed can refer to a pixel that needs to be traversed. The client can determine traversal interval information. At least one pixel to be traversed is obtained by traversing the first labeled image according to the traversal interval information. Labeling path information corresponding to the first labeled image is obtained according to position information of the at least one pixel to be traversed. The traversal interval information can represent the number of rows and columns to be spaced between the pixel corresponding to the current traversal operation and the pixel corresponding to the next traversal operation.

[0094] According to an embodiment of the present disclosure, after determining the pixel to be traversed, the client can determine at least one target labeling pixel from the at least one pixel to be traversed according to the pixel value of the at least one pixel to be traversed. Labeling path information corresponding to the first labeled image is obtained according to position information of the at least one target labeling pixel.

[0095] According to an embodiment of the present disclosure, since the labeling path information is obtained according to the position information of the at least one pixel to be traversed in the first labeled image, the actual size information of the first labeled image is reduced relative to the second actual size information of the second target image, and thus the amount of calculation for determining the labeling path information can be reduced, and the storage speed of the labeling path information is improved.

[0096] According to an embodiment of the present disclosure, obtaining the labeling path information corresponding to the first labeled image according to the position information of the at least one pixel to be traversed can include the following operations.

[0097] At least one target labeling pixel is determined from the at least one pixel to be traversed according to the pixel value of the at least one pixel to be traversed. Labeling path information corresponding to the first labeled image is obtained according to position information of the at least one target labeling pixel.

[0098] According to an embodiment of the present disclosure, for a to-be-traversed pixel in the at least one to-be-traversed pixel, in a case where it is determined that the pixel value of the to-be-traversed pixel is a non-predefined pixel value, it can be considered that the to-be-traversed pixel is a labeled pixel, and thus it can be determined that the to-be-traversed pixel is a target labeled pixel. In a case where it is determined that the pixel value of the to-be-traversed pixel is a predefined pixel value, it can be considered that the to-be-traversed pixel is an unlabeled pixel, and thus it can be determined that the to-be-traversed pixel is a non-target labeled pixel. The predefined pixel value can be configured according to actual business requirements, which is not limited herein.

[0099] According to an embodiment of the present disclosure, the annotation path information corresponding to the first annotation image can be obtained according to the position information of the at least one target labeled pixel.

[0100] According to an embodiment of the present disclosure, since the at least one target labeled pixel is determined from the at least one to-be-traversed pixel according to the pixel value of the at least one to-be-traversed pixel, and the annotation path information is obtained according to the position information of the at least one target labeled pixel, the actual size information of the first annotation image is reduced relative to the second actual size information of the second target image, and thus the calculation amount for determining the annotation path information can be reduced, and the storage speed of the annotation path information is improved.

[0101] According to an embodiment of the present disclosure, determining the at least one target labeled pixel from the at least one to-be-traversed pixel according to the pixel value of the at least one to-be-traversed pixel can include the following operations.

[0102] For a to-be-traversed pixel in the at least one to-be-traversed pixel, in a case where it is determined that the pixel value of the to-be-traversed pixel is a non-predefined pixel value, it is determined that the to-be-traversed pixel is a target labeled pixel.

[0103] According to an embodiment of the present disclosure, in a case where it is determined that the pixel value of the to-be-traversed pixel is a non-predefined pixel value, it can be considered that the to-be-traversed pixel is a labeled pixel, and thus the to-be-traversed pixel can be determined as a target labeled pixel.

[0104] According to an embodiment of the present disclosure, in a case where it is determined that the pixel value of the to-be-traversed pixel is a predefined pixel value, it can be considered that the to-be-traversed pixel is an unlabeled pixel, and thus the to-be-traversed pixel can be determined as a non-target labeled pixel.

[0105] According to an embodiment of the present disclosure, the form of the pixel value can be configured according to actual business requirements, which is not limited herein. For example, the form of the pixel value can be RGB. Alternatively, the form of the pixel value can be RGBA.

[0106] Based on the predetermined ratio information, the position information to be traversed is determined. The first labeled image is traversed according to the position information to be traversed, obtaining at least one pixel in the first labeled image to be traversed.

[0107] According to embodiments of this disclosure, the predetermined ratio information may include a first predetermined ratio and a second predetermined ratio.

[0108] According to embodiments of this disclosure, determining at least one target annotation pixel from at least one pixel to be traversed based on the pixel value of at least one pixel to be traversed may include the following operations.

[0109] In determining the p-th element in the first labeled image and the second target image i / R1，j / R2 The pixel to be traversed corresponds to the p-th pixel. i，j If the pixel value of the pixel to be traversed is not a predetermined pixel value, determine the p-th pixel in the first labeled image. i，j The pixels to be traversed are the target labeled pixels.

[0110] According to embodiments of this disclosure, in the second target image, the p-th... i / R1，j / R2 The pixel to be traversed can represent the pixel in the i / R1-th row and j / R2-th column of the second target image. The p-th pixel in the first labeled image... i，j The pixel to be traversed can represent the pixel in the i-th row and j-th column of the first labeled image. The p-th pixel in the second target image... i / R1，j / R2 The pixel to be traversed and the p-th pixel in the first labeled image i，j Each pixel to be traversed corresponds to a given pixel. R1 can represent a first predetermined ratio. R2 can represent a second predetermined ratio. i can be an integer greater than or equal to 1 and less than or equal to 1. I can represent the total number of rows in the first labeled image. I can be an integer greater than or equal to 1. j can be an integer greater than or equal to 1 and less than or equal to J. J can represent the total number of columns in the first labeled image. J can be an integer greater than or equal to 1.

[0111] According to embodiments of this disclosure, the first predetermined ratio can be greater than 0 and less than 1. The second predetermined ratio can be greater than 0 and less than 1. The values ​​of I and J can be configured according to actual business needs and are not limited here. i∈{1,2,……,I-1,I}. j∈{1,2,……,J-1,J}. i / R1∈{1,2,……,I-1 / R1,I / R1}. j / R2∈{1,2,……,J-1 / R2,J / R2}.

[0112] According to embodiments of this disclosure, the p-th element in the first annotated image can be determined. i，j If the pixel value of the p-th pixel to be traversed is not a predetermined pixel value, it can be explained that the p-th pixel... i，jThe pixel to be traversed is a pixel that is not labeled, and thus the p i，j th pixel to be traversed can be determined as a non-target labeled pixel.

[0113] According to an embodiment of the present disclosure, in a case where the pixel value of the p i，j th pixel to be traversed is a predetermined pixel value, it can be indicated that the p i，j th pixel to be traversed is a pixel that is not labeled, and thus the p i，j th pixel to be traversed can be determined as a non-target labeled pixel.

[0114] For example, the second actual size information of the second target image includes a third width and a third height. The second actual size information of the second target image is "MxN". "M" can represent the third width, that is, the total number of rows included in the second target image. "N" can represent the third height, that is, the total number of columns included in the second target image. The third actual size information of the first labeling interface includes a fourth width and a fourth height. The third actual size information of the first labeling interface is "UxV". "U" can represent the fourth width, that is, the total number of rows included in the first labeling interface. "V" can represent the fourth height, that is, the total number of columns included in the first labeling interface. The tenth actual size information of the first labeling image can include a tenth width and a tenth height. The tenth actual size information of the first labeling image matches the third actual size information. The tenth actual size information of the first labeling image is "UxV". In addition, "U" can also represent the tenth width, that is, the total number of rows included in the first labeling image. "V" can also represent the tenth height, that is, the total number of columns included in the first labeling image.

[0115] The predetermined ratio information can include a first predetermined ratio and a second predetermined ratio. The first predetermined ratio can be determined according to the third width and the fourth width. For example, the first predetermined ratio R1 can be U / M. The second predetermined ratio can be determined according to the third height and the fourth height. For example, the second predetermined ratio R2 can be V / N.

[0116] Based on the above, the number of pixels to be traversed in the first labeling image determined by the image processing method of an embodiment of the present disclosure is "YxMxNxR1xR2", that is, "YxUxV". The number of pixels to be traversed required for directly traversing the labeling image corresponding to the second target image is "YxMxN". "Y" can represent the number of labeling regions. Y can be an integer greater than or equal to 1.

[0117] Since one calculation is needed for each pixel to be traversed, the number of calculations needed for the traversal operation using the image processing method of the embodiments of the present disclosure is "YXMXNXR1XR2", and the number of calculations needed for the traversal operation directly on the labeled image corresponding to the second target image is "YXMXN". As can be seen, the traversal operation using the image processing method of the embodiments of the present disclosure greatly reduces the amount of calculation, thereby improving the storage speed of the labeled path information, thereby alleviating the lag problem of the client, such as a browser.

[0118] For example, in the case of Y=10, M=N=10000, U=V=500, R1=R2=0.05. Thus, the number of calculations needed for the traversal operation using the image processing method of the embodiments of the present disclosure is 2.5x10 6 times. The number of calculations needed for the traversal operation directly on the labeled image corresponding to the second target image is 10 9 times. As can be seen, the image processing method of the embodiments of the present disclosure can compress the amount of calculation by 400 times, improve the storage speed of the labeled path information, and thereby alleviate the lag problem of the client, such as a browser.

[0119] Figure 4 An example schematic diagram of a pixel to be traversed according to an embodiment of the present disclosure is schematically shown.

[0120] As Figure 4 shown in 400, the image 401 can represent the image obtained by the labeling operation on the second target image, that is, the image 401 is the fourth labeled image 401. The image 402 can represent the image obtained by labeling the first target image, that is, the image 402 is the first labeled image 402. Figure 4 The labeled pixels in the middle rectangle grid.

[0121] The second actual size information of the second target image is "16x16", that is, the eleventh actual size information of the fourth labeled image is "16x16". The fifth actual size information of the first target image is "8x8", that is, the tenth actual size information of the first labeled image is "8x8".

[0122] The number of calculations needed for the traversal operation using the image processing method of the embodiments of the present disclosure is 64 times. The number of calculations needed for the traversal operation directly on the fourth labeled image 401 is 256 times. As can be seen, the image processing method of the embodiments of the present disclosure can compress the amount of calculation by 4 times, improve the storage speed of the labeled path information, and thereby alleviate the lag problem of the client, such as a browser.

[0123] According to an embodiment of the present disclosure, the above image processing method can further include the following operations.

[0124] In response to receiving the image recovery information from the server, the second annotation image is generated according to the image recovery information.

[0125] According to an embodiment of the present disclosure, the sixth actual size information and the sixth display size information of the second annotation image are the same. The sixth actual size information and the sixth display size information can both match the second actual size information of the second target image. In response to receiving the image recovery information from the server, the second annotation image can be generated according to the image recovery information.

[0126] According to an embodiment of the present disclosure, since the server generates the second annotation image corresponding to the first annotation image according to the predetermined ratio information and the annotation path information, the client does not need to participate in the calculation process of the second annotation image, thereby releasing the memory space of the client such as a browser and alleviating the problem of the client such as a browser freezing.

[0127] According to an embodiment of the present disclosure, the image processing method described above can further include the following operations.

[0128] In response to detecting the first display instruction, the first display image is displayed on the first annotation interface.

[0129] According to an embodiment of the present disclosure, the first display image can include the second annotation image and the second target image. The second annotation image can be located on an annotation layer of the first annotation interface. The second target image can be located on an image layer of the first annotation interface.

[0130] According to an embodiment of the present disclosure, the seventh actual size information and the seventh display size information of the first display image are the same. The seventh actual size information and the seventh display size information of the first display image can both match the second actual size information of the second target image.

[0131] According to an embodiment of the present disclosure, the first display instruction can refer to an instruction for displaying the first display image. The first display instruction can be generated by the client in a case where it is detected that the first display operation is triggered. Alternatively, the first display instruction can also be generated by the client in a case where it is detected that the first predetermined routine is triggered.

[0132] According to an embodiment of the present disclosure, the client can display the second annotation image on the annotation layer of the first annotation interface and display the second target image on the image layer of the first annotation interface in a case where the first display instruction is detected, so as to display the first display image by cooperation of the annotation layer of the first annotation interface and the image layer of the first annotation interface.

[0133] According to an embodiment of the present disclosure, the first target image is displayed on an image layer of the first annotation interface, the first annotation image corresponding to the first target image is displayed on an annotation layer of the first annotation interface, and the first display image is displayed through cooperation of the annotation layer of the first annotation interface and the image layer of the first annotation interface, so that the first display image is quickly rendered.

[0134] According to an embodiment of the present disclosure, the image processing method can further include the following operations.

[0135] In response to receiving the image recovery information from the server, a third annotation image is generated according to the image recovery information and fourth actual size information of the second annotation interface. A third target image is generated according to the second target image and the fourth actual size information.

[0136] According to an embodiment of the present disclosure, when the annotation interface is the second annotation interface, the second annotation interface can have the same functions and structures as the annotation interface, which will not be described here.

[0137] According to an embodiment of the present disclosure, the eighth display size information and the eighth actual size information of the third annotation image can be the same. The eighth display size information and the eighth actual size information of the third annotation image can both match the fourth actual size information. The ninth display size information and the ninth actual size information of the third target image can be the same. The ninth display size information and the ninth actual size information of the third target image can both match the fourth actual size information.

[0138] According to an embodiment of the present disclosure, the second annotation image can be generated according to the image recovery information. The sixth actual size information of the second annotation image is adjusted to the fourth actual size information to generate the third annotation image. The second actual size information of the second target image can be adjusted to the fourth display information to obtain the third target image.

[0139] According to an embodiment of the present disclosure, the image processing method can further include the following operations.

[0140] In response to detecting the second display instruction, a second display image is displayed on the second annotation interface.

[0141] According to an embodiment of the present disclosure, the second display image can include a third annotation image and a third target image. The third annotation image can be located on an annotation layer of the second annotation interface. The third target image can be located on an image layer of the second annotation interface.

[0142] According to an embodiment of the present disclosure, the second display instruction can be an instruction for displaying the second display image. The second display instruction can be generated by the client when it is detected that the second display operation is triggered. Alternatively, the second display instruction can also be generated by the client when it is detected that the second predetermined routine is triggered.

[0143] According to an embodiment of the present disclosure, the client can display the third annotation image on the annotation layer of the second annotation interface and display the third target image on the image layer of the second annotation interface to display the second display image cooperatively through the annotation layer of the second annotation interface and the image layer of the second annotation interface when the second display instruction is detected.

[0144] According to an embodiment of the present disclosure, the third target image is displayed on the image layer of the second annotation interface, and the third annotation image corresponding to the third target image is displayed on the annotation layer of the second annotation interface, so that the second display image is displayed cooperatively through the annotation layer of the second annotation interface and the image layer of the second annotation interface, thereby realizing fast rendering of the second display image and adapting to the actual size information of the second annotation interface.

[0145] Figure 5 A flowchart of an image processing method according to another embodiment of the present disclosure is schematically shown.

[0146] As shown in Figure 5 The method 500 includes operation S510.

[0147] At operation S510, in response to receiving the predetermined ratio information and the annotation path information from the client, image recovery information corresponding to the first annotation image is determined according to the predetermined ratio information and the annotation path information.

[0148] According to an embodiment of the present disclosure, the annotation path information can be determined by the client according to the predetermined ratio information and the first annotation image. The first annotation image can be displayed on the first annotation interface by the client in response to detecting the annotation operation on the first target image. The first annotation image can include an annotation region. The annotation region can include at least one annotation pixel. The annotation path information can include position information of the at least one annotation pixel.

[0149] According to an embodiment of the present disclosure, the first display size information of the annotation region can be determined according to the predetermined ratio information and the first actual size information of the annotation region. The predetermined ratio information can include a predetermined ratio greater than 0 and less than 1.

[0150] According to an embodiment of the present disclosure, the server can generate the image recovery information according to the predetermined ratio information and the annotation path information. The image recovery information can include position information of a target annotation pixel in an annotation image corresponding to the second target image. The annotation image corresponding to the second target image can refer to an actual annotation image.

[0151] According to an embodiment of the present disclosure, since the first actual size information of the annotation region included in the first annotation image is determined according to the predetermined ratio information and the second actual size information of the second target image corresponding to the first target image, the predetermined ratio information includes a predetermined ratio greater than 0 and less than 1, thus reducing the annotation region of the first target image, and further improving the annotation efficiency for the first target image. In addition, since the annotation path information corresponding to the first annotation image is determined according to the first annotation image, the amount of calculation for determining the annotation path information can be reduced, and the storage speed of the annotation path information is improved. Since the server generates the image recovery information corresponding to the first annotation image according to the predetermined ratio information and the annotation path information without the participation of the client, the memory space of the client such as a browser is released, and the problem of client such as browser lag is alleviated.

[0152] According to an embodiment of the present disclosure, the predetermined ratio information can be determined according to the second actual size information of the second target image corresponding to the first target image and the third actual size information of the first annotation interface.

[0153] According to an embodiment of the present disclosure, the image processing method described above can further include the following operations.

[0154] The image recovery information is sent to the client, so that the client generates a second annotation image according to the image recovery information.

[0155] According to an embodiment of the present disclosure, the image processing method described above can further include the following operations.

[0156] The image recovery information is sent to the client, so that the client generates a third annotation image according to the image recovery information and the fourth actual size information of the second annotation interface. The third target image is generated according to the second target image and the fourth actual size information.

[0157] According to an embodiment of the present disclosure, the image processing method described above can be executed by a server. The server can be a service 105 in Figure 1 .

[0158] Figure 6 The signaling diagram of the image processing method according to an embodiment of the present disclosure is schematically shown.

[0159] As shown in Figure 6 , the method 600 includes operations S601-S610.

[0160] In operation S601, the client adjusts the second actual size information of the second target image to the third actual size information of the first annotation interface to obtain the first target image.

[0161] At operation S602, the client determines the predetermined ratio information according to the second actual size information of the second target image and the third actual size information of the first annotation interface.

[0162] At operation S603, the client displays the first annotation image on the first annotation interface in response to detecting the annotation operation on the first target image.

[0163] At operation S604, the client determines at least one pixel to be traversed in the first annotation image.

[0164] At operation S605, the client determines at least one target annotation pixel from the at least one pixel to be traversed according to a pixel value of the at least one pixel to be traversed.

[0165] At operation S606, the client obtains annotation path information corresponding to the first annotation image according to position information of the at least one target annotation pixel.

[0166] At operation S607, the client sends the predetermined ratio information and the annotation path information to the server.

[0167] At operation S608, the server determines image restoration information corresponding to the first annotation image according to the predetermined ratio information and the annotation path information.

[0168] At operation S609, the server sends the image restoration information to the client.

[0169] At operation S610, the client generates a second annotation image corresponding to the first annotation image according to the image restoration information.

[0170] Figure 7 A flowchart schematically showing a training method of a visual task processing model according to an embodiment of the present disclosure is shown.

[0171] As shown in Figure 7 , the method 700 includes operations S710-S720.

[0172] At operation S710, a display image is obtained.

[0173] At operation S720, a deep learning model is trained using the display image to obtain a visual task processing model.

[0174] According to an embodiment of the present disclosure, the display image can include one of a first display image and a second display image. The first display image can include a second annotation image and a second target image. The second display image can include a third annotation image and a third target image. The first display image can be obtained according to the image processing method described in an embodiment of the present disclosure. The second display image can be obtained according to the image processing method described in an embodiment of the present disclosure.

[0175] According to an embodiment of the present disclosure, the model structure of the deep learning model can be configured according to actual business requirements, which is not limited here and can be implemented as long as the visual task is implemented.

[0176] According to an embodiment of the present disclosure, according to the type of the visual task, the visual task can include at least one of a visual recognition task, a visual detection task, a visual segmentation task, and a visual classification task, etc. According to the scene of the visual task, the visual task can include at least one of a visual task in an industrial quality inspection scene, a visual task in a medical scene, a visual task in a resource recommendation scene, a visual task in an autonomous driving scene, a visual task in a meta-universe scene, a visual task in an augmented reality scene, a visual task in a virtual reality scene, and a visual task in a mixed reality scene, etc. The industrial quality inspection scene can include at least one of an intelligent instrument quality inspection scene, a part quality inspection scene, a photovoltaic electroluminescence quality inspection scene, and a chemical fiber cake appearance defect quality inspection scene, etc.

[0177] According to an embodiment of the present disclosure, the display image can be used as a sample image. Training the deep learning model using the display image to obtain the visual task processing model can include: inputting the sample image into the deep learning model to obtain sample visual processing information of the sample image. Based on the loss function, the sample visual processing information and the sample label information of the sample image are used to obtain a loss function value. The model parameters of the deep learning model are adjusted according to the loss function value until a predetermined end condition is met. The deep learning model obtained when the predetermined end condition is met is determined as the visual task processing model. The predetermined end condition can include at least one of loss function value convergence and training round reaching the maximum training round.

[0178] For example, in the case of a visual task being a visual task in an intelligent instrument quality inspection scene and the display image being a sample instrument image. Training the deep learning model using the display image to obtain the visual task processing model can include: determining a sample dial image from the sample instrument image according to sample detection information obtained by detecting the sample instrument image of the pointer instrument. The sample dial image is subjected to key point detection to obtain sample scale detection information. The sample scale detection information includes sample position information of at least one sample key point. The sample key point can represent a point of a target sample scale. The target sample scale has at least one sample key point corresponding to the target sample scale. The target sample scale is a scale with a number. Sample pointer recognition information is obtained according to sample image segmentation information obtained by image segmentation of the sample dial image. Sample instrument number recognition information is obtained according to the sample scale detection information, the sample pointer recognition information, and sample scale number information corresponding to the sample scale detection information. The deep learning model is trained according to the sample instrument number recognition information and sample instrument number label information to obtain the visual task processing model.

[0179] According to the embodiment of the present disclosure, since the display image is obtained by using the image processing method of the embodiment of the present disclosure, the annotation efficiency for the second target image is improved. In addition, the display image is used to train the deep learning model, and the model training speed is improved.

[0180] Figure 8 A flowchart of a visual task processing method according to an embodiment of the present disclosure is schematically shown.

[0181] As shown in Figure 8 , the method 800 includes operations S810-S820.

[0182] In operation S810, a visual image to be processed is obtained.

[0183] In operation S820, the visual image to be processed is processed by using a visual task processing model to obtain visual processing information.

[0184] According to the embodiment of the present disclosure, the visual task processing model can be obtained by using the training method of the visual task processing model according to the embodiment of the present disclosure.

[0185] Figure 9 An example schematic diagram of a visual task processing method according to an embodiment of the present disclosure is schematically shown.

[0186] As shown in Figure 9 , in 900, the visual image to be processed 901 can be input into the visual task processing model 902 to obtain the visual processing information 903.

[0187] The above is only an exemplary embodiment, but is not limited thereto, and other image processing methods, training methods of visual task processing models, and visual task processing methods known in the art can also be included, as long as they can improve the annotation efficiency and improve the storage speed of the annotation path information.

[0188] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in the technical solution comply with the relevant legal regulations and do not violate public order and good customs.

[0189] Figure 10 A block diagram of an image processing apparatus according to an embodiment of the present disclosure is schematically shown.

[0190] As shown in Figure 10 , the image processing apparatus 1000 can include a first display module 1010, a first determination module 1020, and a first sending module 1030.

[0191] The first display module 1010 is configured to display a first annotation image on the first annotation interface in response to detecting the annotation operation on the first target image. The first annotation image comprises an annotation region. The annotation region comprises at least one annotation pixel. First display size information of the annotation region is determined according to predetermined ratio information and first actual size information of the annotation region. The predetermined ratio information comprises a predetermined ratio greater than 0 and less than 1.

[0192] The first determination module 1020 is configured to determine annotation path information corresponding to the first annotation image according to the first annotation image. The annotation path information comprises position information of the at least one annotation pixel.

[0193] The first sending module 1030 is configured to send the predetermined ratio information and the annotation path information to a server, so that the server obtains image recovery information corresponding to the first annotation image according to the predetermined ratio information and the annotation path information.

[0194] According to an embodiment of the present disclosure, the predetermined ratio information is determined according to second actual size information of a second target image corresponding to the first target image and third actual size information of the first annotation interface.

[0195] According to an embodiment of the present disclosure, the image processing apparatus 1000 described above can further comprise an adjustment module.

[0196] The adjustment module is configured to adjust the second actual size information of the second target image to the third actual size information to obtain the first target image.

[0197] According to an embodiment of the present disclosure, the first display module 1010 can comprise a first determination sub-module, a second determination sub-module, a generation sub-module and a display sub-module.

[0198] The first determination sub-module is configured to determine to trigger the annotation operation on the first target image in response to detecting a mouse start event on the first target image.

[0199] The second determination sub-module is configured to determine to end the annotation operation on the first target image in response to detecting a mouse end event on the first target image.

[0200] The generation sub-module is configured to generate the first annotation image according to the annotation region generated in a process from triggering the annotation operation to ending the annotation operation in response to detecting the end of the annotation operation on the first target image.

[0201] The display sub-module is configured to display the first annotation image on an annotation layer. The annotation layer is located on the first annotation interface.

[0202] According to an embodiment of the present disclosure, second display size information of the annotation layer matches the third actual size information.

[0203] According to an embodiment of the present disclosure, the first target image is located in an image layer. The image layer is located in the first annotation interface. The annotation layer is located above the image layer. The third display size information of the image layer matches the third actual size information.

[0204] According to an embodiment of the present disclosure, the first determining module 1020 can include a third determining sub-module and a first obtaining sub-module.

[0205] The third determining sub-module is configured to determine at least one to-be-traversed pixel in the first annotation image.

[0206] The first obtaining sub-module is configured to obtain annotation path information corresponding to the first annotation image according to position information of the at least one to-be-traversed pixel.

[0207] According to an embodiment of the present disclosure, the number of to-be-traversed pixels is less than or equal to the total number of pixels. The number of to-be-traversed pixels represents the sum of the number of at least one to-be-traversed pixel. The total number of pixels represents the sum of the number of pixels in the first annotation image.

[0208] According to an embodiment of the present disclosure, the first obtaining sub-module can include a determining unit and an obtaining unit.

[0209] The determining unit is configured to determine at least one target annotation pixel from the at least one to-be-traversed pixel according to a pixel value of the at least one to-be-traversed pixel.

[0210] The obtaining unit is configured to obtain annotation path information corresponding to the first annotation image according to position information of the at least one target annotation pixel.

[0211] According to an embodiment of the present disclosure, the determining unit can include a determining sub-unit.

[0212] The determining sub-unit is configured to determine, for a to-be-traversed pixel in the at least one to-be-traversed pixel, that the to-be-traversed pixel is the target annotation pixel in a case where the pixel value of the to-be-traversed pixel is a non-predefined pixel value.

[0213] According to an embodiment of the present disclosure, the image processing apparatus 1000 described above can further include a first generating module.

[0214] The first generating module is configured to, in response to receiving image recovery information from a server, generate a second annotation image according to the image recovery information.

[0215] According to an embodiment of the present disclosure, the image processing apparatus 1000 described above can further include a second display module.

[0216] The second display module is configured to display a first display image in the first annotation interface in response to detecting the first display instruction. The first display image includes a second annotation image and a second target image. The second annotation image is located in an annotation layer of the first annotation interface. The second target image is located in an image layer of the first annotation interface.

[0217] According to an embodiment of the present disclosure, the image processing apparatus 1000 described above can further include a second generation module and a third generation module.

[0218] In response to receiving image recovery information from the server,

[0219] The second generation module is configured to generate a third annotation image according to the image recovery information and fourth actual size information of the second annotation interface.

[0220] The third generation module is configured to generate a third target image according to the second target image and the fourth actual size information.

[0221] According to an embodiment of the present disclosure, the image processing apparatus 1000 described above can further include a third display module.

[0222] The third display module is configured to display a second display image in the second annotation interface in response to detecting a second display instruction. The second display image includes a third annotation image and a third target image. The third annotation image is located in an annotation layer of the second annotation interface. The third target image is located in an image layer of the second annotation interface.

[0223] According to an embodiment of the present disclosure, the annotation operation includes a smearing annotation operation.

[0224] Figure 11 A block diagram of an image processing apparatus according to another embodiment of the present disclosure is schematically shown.

[0225] As Figure 11 shown, the image processing apparatus 1100 can include a second determination module 1110.

[0226] The second determination module 1110 is configured to determine image recovery information corresponding to the first annotation image according to the predetermined ratio information and the annotation path information in response to receiving the predetermined ratio information and the annotation path information from the client.

[0227] According to an embodiment of the present disclosure, the annotation path information is determined by the client according to the first annotation image. The first annotation image is displayed in the first annotation interface by the client in response to detecting an annotation operation on the first target image. The first annotation image includes an annotation region, and the annotation region includes at least one annotation pixel. The annotation path information includes position information of the at least one annotation pixel.

[0228] According to an embodiment of the present disclosure, the first display size information of the labeling region is determined according to predetermined ratio information and first actual size information of the labeling region. The predetermined ratio information includes a predetermined ratio greater than 0 and less than 1.

[0229] According to an embodiment of the present disclosure, the predetermined ratio information is determined according to second actual size information of a second target image corresponding to the first target image and third actual size information of the first labeling interface.

[0230] According to an embodiment of the present disclosure, the image processing apparatus 1100 can further include a second sending module.

[0231] The second sending module is configured to send image recovery information to the client, so that the client generates a second labeled image according to the image recovery information.

[0232] According to an embodiment of the present disclosure, the image processing apparatus 1100 can further include a third sending module.

[0233] The third sending module is configured to send image recovery information to the client, so that the client generates a third labeled image according to the image recovery information and fourth actual size information of the second labeling interface, and generates a third target image according to the second target image and the fourth actual size information.

[0234] Figure 12 A block diagram of a training apparatus of a visual task processing model according to an embodiment of the present disclosure is schematically shown.

[0235] As shown in Figure 12 The training apparatus 1200 of the visual task processing model can include a first obtaining module 1210 and a first obtaining module 1220.

[0236] The first obtaining module 1210 is configured to obtain a display image.

[0237] The first obtaining module 1220 is configured to train a deep learning model using the display image to obtain a visual task processing model.

[0238] According to an embodiment of the present disclosure, the display image includes one of a first display image and a second display image. The first display image includes a second labeled image and a second target image. The second display image includes a third labeled image and a third target image. The second labeled image is obtained by the image processing apparatus according to an embodiment of the present disclosure. The second display image is obtained by the image processing apparatus according to an embodiment of the present disclosure.

[0239] Figure 13 A block diagram of a visual task processing apparatus according to an embodiment of the present disclosure is schematically shown.

[0240] As shown in Figure 13As shown, the visual task processing apparatus 1300 can include a second acquisition module 1310 and a second obtaining module 1320.

[0241] The second acquisition module 1310 is configured to acquire a visual image to be processed.

[0242] The second obtaining module 1320 is configured to process the visual image to be processed by using a visual task processing model to obtain visual processing information.

[0243] According to an embodiment of the present disclosure, the visual task processing model is trained by using a training apparatus of the visual task processing model according to an embodiment of the present disclosure.

[0244] According to an embodiment of the present disclosure, the present disclosure further provides an electronic device, a readable storage medium and a computer program product.

[0245] According to an embodiment of the present disclosure, an electronic device includes a plurality of processors, and a memory connected with the plurality of processors in communication; wherein the memory stores instructions executable by the plurality of processors, and the instructions are executed by the plurality of processors to enable the plurality of processors to perform the method described above.

[0246] According to an embodiment of the present disclosure, a non-transitory computer readable storage medium stores computer instructions, wherein the computer instructions are used to enable a computer to perform the method described above.

[0247] According to an embodiment of the present disclosure, a computer program product includes a computer program, and the computer program, when executed by a processor, implements the method described above.

[0248] Figure 14 A block diagram schematically shows an electronic device suitable for implementing the image processing method, the training method of the visual task processing model and the visual task processing method according to an embodiment of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not intended to limit the implementations of the present disclosure described and / or claimed in this document.

[0249] As Figure 14As shown, the electronic device 1400 includes a computing unit 1401 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 1402 or a computer program loaded into a random access memory (RAM) 1403 from a storage unit 1408. Various programs and data required for the operation of the electronic device 1400 can also be stored in the RAM 1403. The computing unit 1401, the ROM 1402, and the RAM 1403 are connected to each other through a bus 1404. An input / output (I / O) interface 1405 is also connected to the bus 1404.

[0250] A plurality of components in the electronic device 1400 are connected to the I / O interface 1405, including an input unit 1406 such as a keyboard, a mouse, and the like, an output unit 1407 such as various types of displays, a speaker, and the like, a storage unit 1408 such as a magnetic disk, an optical disk, and the like, and a communication unit 1409 such as a network card, a modem, a wireless communication transceiver, and the like. The communication unit 1409 allows the electronic device 1400 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0251] The computing unit 1401 can be various general and / or special-purpose processing components having processing and computing capabilities. Some examples of the computing unit 1401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, and the like. The computing unit 1401 performs various methods and processes described above, such as the image processing method, the training method of the visual task processing model, and the visual task processing method. For example, in some embodiments, the image processing method, the training method of the visual task processing model, and the visual task processing method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 1408. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 1400 via the ROM 1402 and / or the communication unit 1409. When the computer program is loaded into the RAM 1403 and executed by the computing unit 1401, one or more steps of the image processing method, the training method of the visual task processing model, and the visual task processing method described above can be performed. Alternatively, in other embodiments, the computing unit 1401 can be configured to perform the image processing method, the training method of the visual task processing model, and the visual task processing method by any other appropriate means, such as by means of firmware.

[0252] The various embodiments of the systems and techniques described above can be implemented in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a complex programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including multiple programmable processors, which can be special or general purpose, coupled by a bus, that can receive data and instructions from, and transmit data and instructions to, a storage system, multiple input devices, and multiple output devices.

[0253] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces the functions / operations specified in the flowcharts and / or the block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package, and partially on a remote machine or a server.

[0254] In the context of the present disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0255] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, 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, speech, or tactile input.

[0256] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0257] The computer system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server can arise by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0258] It should be understood that various forms of flow shown above can be used, with steps reordered, added, or removed. For example, the steps recited in the present disclosure can be performed in parallel, in series, or in a different order, without limitation herein, so long as the desired results of the technology of the present disclosure are achieved.

[0259] The specific embodiments described above are not intended to be limiting, and persons skilled in the art will appreciate that various modifications, combinations, sub-combinations and alternatives can be made to the specific embodiments without departing from the spirit and principles of the disclosure. Accordingly, the disclosure is not limited to the specific embodiments described above.

Claims

1. An image processing method, comprising: In response to detecting a labeling operation for a first target image, a first labeled image is displayed on a first labeling interface, wherein the first labeled image includes a labeling region, the labeling region includes at least one labeled pixel, and the first display size information of the labeling region is determined based on a predetermined ratio information and the first actual size information of the labeling region, wherein the predetermined ratio information includes a predetermined ratio greater than 0 and less than 1; Based on the first labeled image, annotation path information corresponding to the first labeled image is determined, wherein the annotation path information includes the position information of the at least one labeled pixel; and The predetermined ratio information and the annotation path information are sent to the server so that the server can obtain image restoration information corresponding to the first annotated image based on the predetermined ratio information and the annotation path information.

2. The method according to claim 1, wherein, The predetermined ratio information is determined based on the second actual size information of the second target image corresponding to the first target image and the third actual size information of the first annotation interface.

3. The method according to claim 2, further comprising: The second actual size information of the second target image is adjusted to the third actual size information to obtain the first target image.

4. The method according to any one of claims 1 to 3, wherein, The step of displaying the first annotated image on the first annotation interface in response to detecting an annotation operation on the first target image includes: In response to detecting a mouse activation event for the first target image, it is determined that a annotation operation for the first target image is triggered; In response to detecting a mouse-end event on the first target image, determine to terminate the annotation operation on the first target image; and In response to detecting the end of the annotation operation for the first target image, the first annotated image is generated based on the annotation regions generated during the period from the triggering of the annotation operation to the end of the annotation operation; The first labeled image is displayed in a label layer, wherein the label layer is located on the first label interface.

5. The method according to claim 4, wherein, The second display size information of the annotation layer matches the third actual size information of the first annotation interface.

6. The method according to any one of claims 3 to 5, wherein, The first target image is located in the image layer, the image layer is located in the first annotation interface, the annotation layer is located above the image layer, and the third display size information of the image layer matches the third actual size information of the first annotation interface.

7. The method according to any one of claims 1 to 6, wherein, The step of determining the annotation path information corresponding to the first annotation image based on the first annotation image includes: Determine at least one pixel to be traversed in the first labeled image; and Based on the position information of the at least one pixel to be traversed, the annotation path information corresponding to the first annotation image is obtained; Wherein, the number of pixels to be traversed is less than or equal to the total number of pixels, the number of pixels to be traversed represents the sum of the number of the at least one pixel to be traversed, and the total number of pixels represents the sum of the number of pixels in the first labeled image.

8. The method according to claim 7, wherein, The step of obtaining the annotation path information corresponding to the first annotation image based on the position information of the at least one pixel to be traversed includes: Based on the pixel value of the at least one pixel to be traversed, determine at least one target labeled pixel from the at least one pixel to be traversed; and Based on the location information of the at least one target labeled pixel, the labeling path information corresponding to the first labeled image is obtained.

9. The method according to claim 8, wherein, The step of determining at least one target labeled pixel from the at least one pixel to be traversed based on the pixel value of the at least one pixel to be traversed includes: For the at least one pixel to be traversed, if it is determined that the pixel value of the pixel to be traversed is not a predetermined pixel value, then the pixel to be traversed is determined to be the target labeled pixel.

10. The method according to any one of claims 1 to 9, further comprising: In response to receiving image recovery information from the server, a second labeled image is generated based on the image recovery information.

11. The method of claim 10, further comprising: In response to detecting a first display instruction, a first display image is displayed on the first annotation interface, wherein the first display image includes a second annotation image and a second target image corresponding to the first target image, the second annotation image is located in the annotation layer of the first annotation interface, and the second target image is located in the image layer of the first annotation interface.

12. The method according to any one of claims 1 to 9, further comprising: In response to receiving image recovery information from the server, A third annotated image is generated based on the image restoration information and the fourth actual size information of the second annotation interface; as well as A third target image is generated based on the second target image corresponding to the first target image and the fourth actual size information.

13. The method of claim 12, further comprising: In response to the detection of a second display instruction, a second display image is displayed on the second annotation interface, wherein the second display image includes the third annotation image and the third target image, the third annotation image is located in the annotation layer of the second annotation interface, and the third target image is located in the image layer of the second annotation interface.

14. The method according to any one of claims 1 to 13, wherein, The annotation operation includes the smearing annotation operation.

15. An image processing method, comprising: In response to receiving predetermined ratio information and annotation path information from the client, image restoration information corresponding to the first annotated image is generated based on the predetermined ratio information and the annotation path information. The annotation path information is determined by the client based on the predetermined ratio information and the first annotation image. The first annotation image is displayed by the client on the first annotation interface in response to detecting an annotation operation on the first target image. The first annotation image includes an annotation area, the annotation area includes at least one annotation pixel, and the annotation path information includes the position information of the at least one annotation pixel. The first display size information of the labeled area is determined based on the predetermined ratio information and the first actual size information of the labeled area, wherein the predetermined ratio information includes a predetermined ratio greater than 0 and less than 1.

16. The method according to claim 15, wherein, The predetermined ratio information is determined based on the second actual size information of the second target image corresponding to the first target image and the third actual size information of the first annotation interface.

17. The method according to claim 15 or 16, further comprising: The image restoration information is sent to the client so that the client can generate a second labeled image based on the image restoration information.

18. The method according to claim 15 or 16, further comprising: The image restoration information is sent to the client so that the client can generate a third labeled image based on the image restoration information and the fourth actual size information of the second labeled interface, and generate a third target image based on the second target image corresponding to the first target image and the fourth actual size information.

19. A method for training a visual task processing model, comprising: Get the displayed image; as well as The deep learning model is trained using the displayed image to obtain the visual task processing model; The displayed image includes one of the following: a first displayed image and a second displayed image, wherein the first displayed image includes a second labeled image and a second target image, the second displayed image includes a third labeled image and a third target image, the first displayed image is obtained by the method according to claim 10 or 11, and the second displayed image is obtained by the method according to claim 12 or 13.

20. A visual task processing method, comprising: Acquire the visual image to be processed; as well as The visual image to be processed is processed using a visual task processing model to obtain visual processing information; The visual task processing model is trained using the method described in claim 19.

21. An image processing apparatus, comprising: A first display module is configured to display a first labeled image on a first labeling interface in response to detecting a labeling operation on a first target image. The first labeled image includes a labeling area, the labeling area includes at least one labeled pixel, and the first display size information of the labeling area is determined based on a predetermined ratio and the first actual size information of the labeling area. The predetermined ratio includes a predetermined ratio greater than 0 and less than 1. The first determining module is configured to determine, based on the predetermined ratio information and the first labeled image, labeling path information corresponding to the first labeled image, wherein the labeling path information includes the position information of the at least one labeled pixel; and The first sending module is used to send the predetermined ratio information and the annotation path information to the server, so that the server can obtain image restoration information corresponding to the first annotated image based on the predetermined ratio information and the annotation path information.

22. The apparatus of claim 21, further comprising: The first generation module is configured to generate a second labeled image in response to receiving image restoration information from the server, based on the image restoration information.

23. The apparatus of claim 21, further comprising: In response to receiving image recovery information from the server, The second generation module is used to generate a third labeled image based on the image recovery information and the fourth actual size information of the second annotation interface; as well as The third generation module is used to generate a third target image based on the second target image corresponding to the first target image and the fourth actual size information.

24. An image processing apparatus, comprising: The second determining module is used to determine image restoration information corresponding to the first labeled image based on the predetermined ratio information and the labeled path information received from the client in response to receiving predetermined ratio information and labeled path information. The annotation path information is determined by the client based on the predetermined ratio information and the first annotation image. The first annotation image is displayed by the client on the first annotation interface in response to detecting an annotation operation on the first target image. The first annotation image includes an annotation area, the annotation area includes at least one annotation pixel, and the annotation path information includes the position information of the at least one annotation pixel. The first display size information of the labeled area is determined based on the predetermined ratio information and the first actual size information of the labeled area, wherein the predetermined ratio information includes a predetermined ratio greater than 0 and less than 1.

25. A training device for a visual task processing model, comprising: The first acquisition module is used to acquire the display image; as well as The first obtaining module is used to train a deep learning model using the displayed image to obtain the visual task processing model. The displayed image includes one of the following: a first displayed image and a second displayed image, wherein the first displayed image includes a second labeled image and a second target image, the second displayed image includes a third labeled image and a third target image, the second labeled image is obtained by the apparatus according to claim 22, and the second displayed image is obtained by the apparatus according to claim 23.

26. A visual task processing device, comprising: The second acquisition module is used to acquire the visual image to be processed; as well as The second obtaining module is used to process the visual image to be processed using a visual task processing model to obtain visual processing information. The visual task processing model is trained using the device according to claim 25.

27. An electronic device comprising: Multiple processors; as well as A memory communicatively connected to the plurality of processors; wherein, The memory stores instructions that can be executed by the plurality of processors to enable the plurality of processors to perform the method of any one of claims 1 to 20.

28. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 20.

29. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 20.

Images