An image instance segmentation-based post-processing optimization method and device

Abstract

The application relates to a post-processing optimization method and device based on image instance segmentation, which comprises the following steps: obtaining one or more segmentation images based on image instance segmentation; extracting one or more contour segmentation lines of a discrete mask in each segmentation image, and judging the hole attribute of each discrete mask according to the pixel difference in each contour segmentation line; filling each discrete mask with the hole attribute; calculating the area and length relationship between each mask based on the mask segmentation image of each segmentation image and the segmentation image with the hole attribute, identifying whether each mask overlaps according to the area and length relationship, filtering the overlapping area, and identifying and filtering the island area of each segmentation image. According to the application, the overlapping, island and hole existing after image instance segmentation are identified and filtered through the overlapping area and length relationship of the mask, so that the segmentation precision of the instance contour is improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of image processing, and particularly relates to a post-processing optimization method and device based on image instance segmentation. BACKGROUND

[0002] In recent years, with the rapid development of automatic driving perception technology, image instance segmentation is increasingly widely used in perception tasks. Due to the introduction of image noise, errors of the segmentation model and other reasons, problems such as overlap (hereinafter referred to as "overlap") of multi-class segmentation mask graphs (hereinafter collectively referred to as "mask") after segmentation, multiple pixel noise (hereinafter referred to as "island") of a single class, and existence of holes (hereinafter referred to as "hole") in the instance inside the segmentation mask graph, etc. make the accuracy of image segmentation low. The traditional image instance segmentation method does not do post-processing and directly outputs the segmentation result, which is difficult to solve the above three problems, resulting in the problem of confusion and error of instance edge pixel class after segmentation. SUMMARY

[0003] In order to solve the problems of overlap, island, hole, etc. after image instance segmentation, and improve the segmentation accuracy of instance contour, in the first aspect of the present application, a post-processing optimization method based on image instance segmentation is provided, comprising: obtaining one or more segmentation images based on image instance segmentation, wherein each segmentation image includes a plurality of masks after segmentation, a confidence map and a mask segmentation image; extracting one or more contour segmentation lines of a discrete mask in each segmentation image, and judging the hole attribute of each discrete mask according to the pixel difference in each contour segmentation line; filling each discrete mask with a hole attribute; calculating the area and length relationship between each mask based on the mask segmentation image of each segmentation image and its corresponding discrete mask with a hole attribute; according to the calculation result of the area and length relationship between each mask, identifying whether each mask overlaps, and filtering the overlapping area; according to the calculation result of the area and length relationship between each mask, identifying and filtering the island area of each segmentation image.

[0004] In some embodiments of the present application, the extraction of one or more contour segmentation lines of a discrete mask in each segmentation image, and the judgment of the hole attribute of each discrete mask according to the pixel difference in each contour segmentation line comprises: counting the maximum pixel gray value and the number of maximum pixel gray value in each contour segmentation line, and judging the hole attribute of each discrete mask according to the difference between the two numbers.

[0005] Further, the calculating the area and length relationship between the masks comprises: obtaining a plurality of single-class masks from the discrete binary map, and taking the intersection of each single-class mask and masks of other classes; calculating the overlapping area between each two single-class masks and the proportion of the mask with the smallest area, and the aspect ratio of each single-class mask; and calculating the area of each single-class discrete mask, the area of each mask and the proportion of the largest mask area.

[0006] In some embodiments of the present application, the identifying whether the masks overlap according to the calculation results of the area and length relationship between the masks comprises: identifying whether the masks overlap according to whether the overlapping area between each two single-class masks and the proportion of the mask with the smallest area, or the aspect ratio of each single-class mask is higher than a threshold value.

[0007] Further, the identifying and filtering the isolated island region of each segmented image according to the calculation results of the area and length relationship between the masks comprises: identifying and filtering the isolated island region of each segmented image according to whether the area of each single-class discrete mask, or the area of each mask and the proportion of the largest mask area is lower than a threshold value.

[0008] In the above-mentioned embodiments, the obtaining one or more segmented images based on image instance segmentation comprises: performing instance segmentation on one or more images by an instance segmentation model to obtain a segmented image of each image after instance segmentation.

[0009] In a second aspect, the present application provides a post-processing optimization device based on image instance segmentation, comprising: an obtaining module configured to obtain one or more segmented images based on image instance segmentation, wherein each segmented image comprises a plurality of masks after segmentation, a confidence map and a mask segmentation image; a hole module configured to determine the hole attribute of each discrete mask according to one or more contour segmentation lines of the discrete mask in each segmented image, and according to the pixel difference in each contour segmentation line; filling each discrete mask with a hole attribute; a coincidence module configured to calculate the area and length relationship between the masks based on the mask segmentation image of each segmented image and its corresponding segmented image with a hole attribute; identifying whether the masks overlap according to the calculation results of the area and length relationship between the masks, and filtering the overlapping region; and an isolated island module configured to identify and filter the isolated island region of each segmented image according to the calculation results of the area and length relationship between the masks.

[0010] In a third aspect, the present application provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the post-processing optimization method based on image instance segmentation provided in the first aspect of the present application.

[0011] In a fourth aspect, the present application provides a computer readable medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the post-processing optimization method based on image instance segmentation provided in the first aspect of the present application.

[0012] The present application has the following beneficial effects:

[0013] The present application relates to a post-processing optimization method and device based on image instance segmentation, the method comprising: obtaining one or more segmentation images based on image instance segmentation; extracting one or more contour segmentation lines of a discrete mask in each segmentation image, and judging the hole attribute of each discrete mask according to the pixel difference in each contour segmentation line; filling each discrete mask with a hole attribute; calculating the area and length relationship between each mask based on the mask segmentation image of each segmentation image and the segmentation image with a hole attribute, and identifying whether each mask overlaps according to the relationship, filtering the overlapping area, and identifying and filtering the island area of each segmentation image. It can be seen that, by sequentially identifying and filtering the holes, overlaps and islands, the judgment result of the next step reuses the judgment result of the previous step, thereby improving the efficiency and accuracy of solving the image instance segmentation hole, overlap and island problem. At the same time, it can filter the problem that the same pixel belongs to different categories, and solve the problem of outputting multiple instance contours due to overlap. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 The figure is a basic flowchart of the post-processing optimization method based on image instance segmentation in some embodiments of the present application;

[0015] Figure 2 The figure is an effect diagram of overlap segmentation in the post-processing optimization method based on image instance segmentation in some embodiments of the present application;

[0016] Figure 3 The figure is an effect diagram of overlap segmentation in the existing instance segmentation method in some embodiments of the present application;

[0017] Figure 4 The figure is an effect diagram of island processing in the post-processing optimization method based on image instance segmentation in some embodiments of the present application;

[0018] Figure 5Effect diagram of island processing of an existing instance segmentation method in some embodiments of the present application;

[0019] Figure 6 Effect diagram of holes in a post-processing optimization method based on image instance segmentation in some embodiments of the present application;

[0020] Figure 7 Effect diagram of holes of an existing instance segmentation method in some embodiments of the present application;

[0021] Figure 8 Structure diagram of a post-processing optimization device based on image instance segmentation in some embodiments of the present application;

[0022] Figure 9 Effect diagram of overlapping segmentation in a post-processing optimization method based on image instance segmentation in some embodiments of the present application;

[0023] Figure 10 Structure diagram of an electronic device in some embodiments of the present application. DETAILED DESCRIPTION

[0024] The principles and features of the present application are described below in conjunction with the accompanying drawings, and the examples are only used to explain the present application and not to limit the scope of the present application.

[0025] Reference Figure 1 In a first aspect of the present application, a post-processing optimization method based on image instance segmentation is provided, comprising: S100. obtaining one or more segmented images after image instance segmentation, wherein each segmented image comprises a plurality of segmented masks, a confidence map and a mask segmented image; S200. extracting one or more contour segmentation lines of a discrete mask in each segmented image, and judging the hole attribute of each discrete mask according to the pixel difference within each contour segmentation line; filling each discrete mask with a hole attribute; S300. calculating the area and length relationship between each mask based on the mask segmented image of each segmented image and its corresponding segmented image with a hole attribute; according to the calculation result of the area and length relationship between each mask, identifying whether each mask overlaps, and filtering the overlapping area; S400. according to the calculation result of the area and length relationship between each mask, identifying and filtering the island area of each segmented image.

[0026] In step S100 of the present disclosure, the obtaining one or more segmented images after image instance segmentation comprises: performing instance segmentation on one or more images by an instance segmentation model to obtain a segmented image of each image after instance segmentation.

[0027] Specifically, the mask2former instance segmentation model is adopted to infer the mask and the confidence map of the segmented single image, and the required mask segmentation image is screened according to the category proportion and the confidence threshold. Optionally, the YOLO series or other instance segmentation models can be used to realize the above instance segmentation step.

[0028] In step S200 of some embodiments of the present application, the one or more contour segmentation lines of each discrete mask in the segmented image are extracted, and the hole attribute of each discrete mask is determined according to the pixel difference in each contour segmentation line, including: counting the maximum pixel gray value and the number of maximum pixel gray value in each contour segmentation line, and determining the hole attribute of each discrete mask according to the difference between the two numbers.

[0029] Specifically, the contour segmentation line of each discrete mask is extracted from the discrete mask obtained by segmentation, the number of pixels 255 and pixels 0 in the contour segmentation line is calculated respectively, and the hole attribute of the mask is obtained according to the difference between the two numbers. Optionally, 255 and 0 are quantized representations of the maximum and minimum pixel gray values, which can be set to other numerical values for binarization according to actual needs.

[0030] Further, the calculation of the area and length relationship between each mask includes: obtaining a plurality of single-class masks from the discrete binary graph, taking the intersection of each single-class mask and the mask of other classes; calculating the overlap area between each two single-class masks and the proportion of the smallest mask area, and the aspect ratio of each single-class mask; calculating the area of each single-class discrete mask, the area of each mask and the proportion of the maximum mask area.

[0031] Specifically, a plurality of single-class masks are obtained from the discrete binary graph, the intersection of these masks and the masks of other classes is taken, the overlap area between the masks and the proportion of the smallest mask area, and the aspect ratio of the masks are calculated. The area of each discrete mask, the proportion of the mask area and the maximum mask area are calculated as the determination filtering conditions.

[0032] In step S300 of some embodiments of the present application, the calculation result of the area and length relationship between each mask is used to identify whether the masks overlap, including: according to the overlap area between each two single-class masks and the proportion of the smallest mask area, or whether the aspect ratio of each single-class mask is higher than the threshold, whether each mask overlaps is identified.

[0033] Specifically, the image processing module calculates the overlapping area, the mask ratio of the overlapping area to the minimum area, and the mask aspect ratio condition, and filters out the overlapping area with a large overlapping ratio and a certain aspect ratio.

[0034] Further, the calculation result according to the area and length relationship between the masks is used to identify and filter the isolated island area of each segmented image, including whether the area of each single-class discrete mask or the proportion of the area of each mask to the maximum mask area is lower than a threshold value.

[0035] The area of each discrete mask calculated by the image processing module, the proportion of the area of each mask to the maximum mask area, is used for judgment, and masks with small area and small proportion are filtered out to solve the "isolated island" problem. The number of pixels 255 and pixels 0 in the contour segmentation line is calculated, the hole attribute of the mask is compared, and the mask belonging to the hole attribute is filled.

[0036] In one embodiment of the present application, the following steps are included:

[0037] Step 1: input the street view image into the instance segmentation model to obtain the segmented image before post-processing;

[0038] The segmented image is input into the hole discrimination module, the contour segmentation line of the discrete mask is calculated, the difference in the number of pixels in the contour segmentation line is calculated, and the hole attribute is assigned to each mask;

[0039] Step 2: input the segmented image before post-processing and the segmented image with hole attribute into the image processing module to calculate the overlapping area between the masks and the mask ratio of the minimum area, and the mask aspect ratio. Calculate the area of each discrete mask, the proportion of the area of each mask to the maximum mask area;

[0040] Step 3: according to the calculation result of the image processing module, input into the overlapping filtering module to identify the overlapping area and filter the mask in the overlapping area;

[0041] Step 4: according to the calculation result of the image processing module, input into the isolated island filtering module to identify the isolated island area and filter the mask in the isolated island area;

[0042] Step 5: according to the calculation result of the image processing module, input into the hole filling module, and according to the hole attribute of the discrete mask, fill the hole area.

[0043] Step 6: according to the calculation result of the image processing module, input into the hole filling module, and according to the hole attribute of the discrete mask, fill the hole area.

[0044] Reference Figures 2 to 6 , which shows the segmentation effect image of the post-processing optimization method based on image instance segmentation and the instance segmentation method without post-processing (referred to as "instance segmentation method"). Among them Figure 2 , Figure 3 are the segmentation effect images of the method and the instance segmentation method on the class overlap problem, respectively. From the effect image, it can be seen that the method can effectively filter the "people" overlapping in the instance "car" compared with the existing instance segmentation method. Figure 4 , Figure 5 are the segmentation effect images of the method and the instance segmentation method on the island problem, respectively. From the effect image, it can be seen that the method can effectively filter the small islands in the instance "lane line" compared with the instance segmentation method. Figure 6 , Figure 7 are the segmentation effect images of the method and the instance segmentation method on the hole problem, respectively. From the effect, it can be seen that the method can effectively fill the holes existing in the instance "road" compared with the instance segmentation.

[0045] Embodiment 2

[0046] Reference Figure 8 , the second aspect of the application, provides a post-processing optimization device 1 based on image instance segmentation, comprising: an acquisition module 11, used for acquiring one or more segmentation images based on image instance segmentation, wherein each segmentation image comprises a plurality of masks, a confidence map and a mask segmentation image after segmentation; a hole module 12, used for judging the hole attribute of each discrete mask according to one or more contour segmentation lines of the discrete mask in each segmentation image, and filling each discrete mask with a hole attribute; a coincidence module 13, used for calculating the area and length relationship between each mask based on the mask segmentation image of each segmentation image and its corresponding segmentation image with a hole attribute; according to the calculation result of the area and length relationship between each mask, identifying whether each mask is coincident, and filtering the coincident area; an island module 14, used for identifying and filtering the island area of each segmentation image according to the calculation result of the area and length relationship between each mask.

[0047] Furthermore, the overlap module includes: a first calculation unit, used to obtain multiple single-category masks through a discrete binary image, and to take the intersection of each single-category mask with the masks of other categories; a second calculation unit, used to calculate the ratio of the overlap area between every two single-category masks to the mask with the smallest area, and the aspect ratio of each single-category mask; and a third calculation unit, used to calculate the area of ​​each single-category discrete mask, and the ratio of the area of ​​each mask to the area of ​​the largest mask.

[0048] refer to Figure 9 In a specific embodiment of the post-processing optimization device 1 based on image instance segmentation described above, it includes:

[0049] Instance segmentation module: The mask2former instance segmentation model is used to infer the mask and confidence map after segmentation of a single image, and the required mask segmentation map is selected according to the category ratio and confidence threshold;

[0050] Hole detection module: Based on the segmented discrete mask, the contour segmentation line of each discrete mask is extracted, and the number of pixels 255 and pixels 0 within the contour segmentation line is calculated respectively. The hole attribute of the mask is obtained according to the difference between the two pixel counts.

[0051] Image processing module: Obtains multiple single-class masks from discrete binary images. Takes the intersection of these masks with masks of other classes, calculates the ratio of the overlapping area between masks to the mask with the smallest area, and the aspect ratio of the masks. Calculates the area of ​​each discrete mask and the ratio of its area to the largest mask area, using these as filtering criteria.

[0052] Overlap filtering module: Filters out overlapping areas with large overlap ratios and aspect ratios within a certain range by taking the overlapping area, the ratio of the overlapping area to the smallest mask area, and the mask aspect ratio criteria calculated by the image processing module.

[0053] Island Filtering Module: Determines the area of ​​each discrete mask calculated by the image processing module and the ratio of the mask area to the largest mask area, and filters out masks with small area and small proportion to solve the "island" problem.

[0054] Hole filling module: By calculating the number of pixels 255 and 0 within the contour dividing line, the hole attribute of the mask is obtained by comparison, and the mask with the hole attribute is filled.

[0055] Example 3

[0056] refer to Figure 10In a third aspect, the present application provides an electronic device comprising: one or more processors; and a memory device storing one or more programs, when executed by the one or more processors, cause the one or more processors to implement the post-processing optimization method based on image instance segmentation of the first aspect of the present application.

[0057] The electronic device 500 can include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 501 that can perform various appropriate actions and processes according to programs stored in a read-only memory (ROM) 502 or loaded into a random access memory (RAM) 503 from a memory device 508. Various programs and data required for the operation of the electronic device 500 are also stored in the RAM 503. The processing device 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0058] In general, the following devices can be connected to the I / O interface 505: input devices 506 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 507 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a memory device 508 including, for example, a hard disk, etc.; and a communication device 509. The communication device 509 can allow the electronic device 500 to communicate wirelessly or by wire with other devices to exchange data. Although Figure 10 The electronic device 500 is shown with various devices, but it should be understood that all of the shown devices are not required to implement or have the electronic device. More or less devices can alternatively be implemented. Figure 10 Each block shown in the flowchart of FIG. 6 can represent a device, or a number of devices, depending on the needs of the implementation.

[0059] In particular, according to embodiments of the present disclosure, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product comprising a computer program carried on a computer readable medium, the computer program containing program code for executing the methods illustrated by the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network by the communication device 509, or installed from the storage device 508, or installed from the ROM 502. When the computer program is executed by the processing device 501, the above-mentioned functions defined in the methods of embodiments of the present disclosure are executed. It should be noted that the computer readable medium described in embodiments of the present disclosure can be a computer readable signal medium or a computer readable storage medium or any combination of the two. The computer readable storage medium may, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination of the above. More specific examples of the computer readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In embodiments of the present disclosure, the computer readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus or device. In embodiments of the present disclosure, the computer readable signal medium can include a data signal carried in a baseband or as part of a carrier wave, in which the computer readable program code is carried. Such a propagated data signal can take a variety of forms, including but not limited to an electromagnetic signal, an optical signal or any suitable combination of the above. The computer readable signal medium can also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate or transport a program for use by or in connection with an instruction execution system, apparatus or device. The program code contained on the computer readable medium can be transmitted by any suitable medium, including but not limited to a wire, cable, RF (radio frequency), or the like, or any suitable combination of the above.

[0060] The computer readable medium described above can be included in the electronic device described above; or can exist separately from the electronic device and be not assembled in the electronic device. The computer readable medium described above carries one or more computer programs, which, when executed by the electronic device, cause the electronic device to:

[0061] Computer program code for carrying out operations of embodiments of the present disclosure can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++, Python, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0062] The computer program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0063] The foregoing is merely illustrative of the principles of the application, and the application should not be limited to such detail. Rather, the spirit and scope of the application are limited solely by the claims.

Claims

1. An image instance segmentation based post-processing optimization method, characterized in that, The method comprises the following steps: obtaining one or more segmented images based on image instance segmentation, wherein each segmented image comprises a plurality of segmented masks, a confidence map and a mask segmented image; extracting one or more contour segmentation lines of each discrete mask in each segmented image, and judging the hole attribute of each discrete mask according to the pixel difference in each contour segmentation line; filling each discrete mask with a hole attribute; calculating the area and length relationship between each mask based on the mask segmented image of each segmented image and the segmented image with a hole attribute corresponding to the mask segmented image; obtaining a plurality of single-class masks through discrete binary images, and taking the intersection of each single-class mask and masks of other classes; calculating the overlapping area between each two single-class masks and the proportion of the smallest mask area, and the aspect ratio of each single-class mask; calculating the area of each single-class discrete mask, the proportion of the area of each mask to the maximum mask area; according to the calculation result of the area and length relationship between each mask, whether the masks overlap is identified, and the overlapping region is filtered; whether the masks overlap is identified according to the calculation result of the area and length relationship between each mask, which comprises: whether the proportion of the overlapping area between each two single-class masks to the smallest mask area, or the aspect ratio of each single-class mask is higher than a threshold value, is identified. According to the calculation result of the area and length relationship between each mask, the isolated island region of each segmented image is identified and filtered.

2. The post-processing optimization method based on image instance segmentation according to claim 1, characterized in that, The extraction of one or more contour segmentation lines of each discrete mask in each segmented image, and the judgment of the hole attribute of each discrete mask according to the pixel difference in each contour segmentation line comprises: statistically counting the maximum pixel gray value in each contour segmentation line and the number of the maximum pixel gray value, and judging the hole attribute of each discrete mask according to the difference between the two numbers.

3. The post-processing optimization method based on image instance segmentation according to claim 1, characterized in that, According to the calculation result of the area and length relationship between each mask, the isolated island region of each segmented image is identified and filtered, which comprises: whether the area of each single-class discrete mask, or the proportion of the area of each mask to the maximum mask area is lower than a threshold value, is identified and filtered to identify and filter the isolated island region of each segmented image.

4. The post-processing optimization method based on image instance segmentation according to any one of claims 1 to 3, characterized in that, The method comprises the following steps: performing instance segmentation on one or more images through an instance segmentation model to obtain a segmented image of each image after instance segmentation.

5. An apparatus for post-processing optimization based on image instance segmentation, characterized in that, The method comprises the following steps: an acquisition module for obtaining one or more segmented images based on image instance segmentation, wherein each segmented image comprises a plurality of segmented masks, a confidence map and a mask segmented image; a hole module for judging the hole attribute of each discrete mask according to one or more contour segmentation lines of each discrete mask in each segmented image, and according to the pixel difference in each contour segmentation line; filling each discrete mask with a hole attribute; The coincidence module is configured to calculate area and length relationships between masks of each segmented image based on the masks and the corresponding segmented images with hole attributes, obtain a plurality of single-class masks from the discrete binary image, and obtain an intersection between each single-class mask and masks of other classes; calculate a coincidence area between each two single-class masks, a proportion of the smallest mask in the area, and an aspect ratio of each single-class mask; calculate an area of each single-class discrete mask, a proportion of the area of each mask in a maximum mask area, and the maximum mask area; and identify whether the masks coincide based on a calculation result of the area and length relationships between the masks, and filter a coincidence region therefrom; and the identification of whether the masks coincide based on the calculation result of the area and length relationships between the masks includes: identifying whether the masks coincide based on whether the coincidence area between each two single-class masks, the proportion of the smallest mask in the area, or the aspect ratio of each single-class mask is higher than a threshold value. The island module is configured to identify and filter an island region of each segmented image based on the calculation result of the area and length relationships between the masks.

6. An electronic device, comprising: one or more processors; a storage device for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement the post-processing optimization method based on image instance segmentation according to any one of claims 1 to 4.

7. A computer readable medium having stored thereon a computer program, wherein, The computer program, when executed by a processor, implements the post-processing optimization method based on image instance segmentation according to any one of claims 1 to 4.

Images