Image segmentation method and device based on multi-task loss function, equipment and medium
By introducing multi-task loss function into the image segmentation network, the accuracy, connectivity and smoothness of the image segmentation model are optimized, and the problem of image segmentation models focusing on accuracy but ignoring connectivity and smoothness in the prior art is solved, which significantly improves the image segmentation effect.
Patent Information
- Application Number
- CN202411974252.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-30
- Publication Date
- 2025-04-29
AI Technical Summary
In the existing image segmentation scheme, the image segmentation model only focuses on accuracy indicators, but ignores the connectivity and smoothness of the image, which leads to the network tending to pay attention to the coarse positioning problem of individual difficult samples, making it difficult to achieve refined positioning of the detected targets.
The accuracy, connectivity and smoothness of the image segmentation network are optimized through the multi-task loss function to build an image segmentation model. Multitasking loss functions include cross entropy loss, Dice loss, connectivity loss, and smoothing loss, and the model is optimized through a combination of these loss functions.
The accuracy, connectivity and smoothness of the image segmentation model are achieved, which significantly improves the performance of the image segmentation model, meets the refined positioning of the detected targets, and improves the image segmentation effect.
Smart Images
Figure CN119904468A_ABST
Abstract
Description
Technical Field
[0001] Embodiments of the present invention relate to the field of image segmentation technology, and in particular, to an image segmentation method, device, equipment and medium based on a multi-task loss function. Background Art
[0002] Image segmentation is an important problem in the field of computer vision, which involves dividing an image into different regions or objects. In recent years, deep learning methods have made significant progress in the field of image segmentation. In deep neural networks, the loss function reflects the performance of the algorithm by calculating the difference between the model's predicted value and the true value. Therefore, it is necessary to find a set of model parameters that minimize the loss function value by training the model to obtain the network model with the best performance. However, in existing image segmentation schemes, the image segmentation model only focuses on the accuracy index, while ignoring the connectivity and smoothness of the image. The cross entropy loss function Softmax Loss is used as the classification loss, and the smooth loss function Smooth L1 Loss is used as the regression loss function. Softmax Loss can only achieve the separability of target categories by finding the best classification surface for each type of target, and the feature discrimination ability is insufficient; the positioning loss value contributed by the difficult samples in Smooth L1 Loss is much greater than that of the regular samples, which causes the network to tend to focus on the coarse positioning problem of individual difficult samples, making it difficult to achieve the refined positioning of the detected target.
[0003] The above problems need to be solved urgently. Summary of the invention
[0004] In order to solve the related technical problems, the present invention provides an image segmentation method, device, equipment and medium based on a multi-task loss function to solve the problems mentioned in the above background technology part.
[0005] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions:
[0006] In a first aspect, an embodiment of the present invention provides an image segmentation method based on a multi-task loss function, the method comprising:
[0007] The accuracy, connectivity and smoothness of the image segmentation network are optimized through a multi-task loss function to construct an image segmentation model.
[0008] receiving a target image acquisition instruction, and acquiring a target image to be segmented according to the target image acquisition instruction;
[0009] The target image to be segmented is input into the image segmentation model for image segmentation.
[0010] As an optional implementation, the multi-task loss function includes but is not limited to cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM.
[0011] As an optional implementation, the accuracy, connectivity and smoothness of the image segmentation network are optimized by a multi-task loss function to construct an image segmentation model, which specifically includes:
[0012] Obtain the image to be segmented;
[0013] Performing image segmentation on the image to be segmented using an image segmentation network to obtain a mask corresponding to the image to be segmented;
[0014] The accuracy, connectivity and smoothness of the image segmentation network are optimized through a multi-task loss function to construct an image segmentation model.
[0015] As an optional implementation, the accuracy, connectivity and smoothness of the image segmentation network are optimized by using a multi-task loss function, including:
[0016] Calculate the cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM of the multi-task loss function; obtain the multi-task loss function according to the cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM.
[0017] As an optional implementation, the accuracy, connectivity and smoothness of the image segmentation network are optimized by using a multi-task loss function, including:
[0018] The accuracy of the image segmentation model is optimized by the cross entropy loss CE; the Dice index of the image segmentation model is optimized by the Dice loss DC; the connectivity of the image segmentation model is optimized by the connectivity loss CC; and the smoothness of the image segmentation model is optimized by the smoothness loss SM.
[0019] In a second aspect, an embodiment of the present invention provides an image segmentation device based on a multi-task loss function, the device comprising:
[0020] An image segmentation model building module is used to optimize the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to build an image segmentation model;
[0021] A target image acquisition module to be segmented, used for receiving a target image acquisition instruction and acquiring the target image to be segmented according to the target image acquisition instruction;
[0022] The image segmentation module is used to input the target image to be segmented into the image segmentation model for image segmentation.
[0023] As an optional implementation, the multi-task loss function includes but is not limited to cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM.
[0024] As an optional implementation, the image segmentation model construction module is specifically used to obtain an image to be segmented; perform image segmentation on the image to be segmented using an image segmentation network to obtain a mask corresponding to the image to be segmented; and optimize the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to construct an image segmentation model.
[0025] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a processor and a memory connected to the processor, wherein program data is stored in the memory, and the processor calls the program data stored in the memory to execute the image segmentation method based on the multi-task loss function provided in the embodiment of the first aspect above.
[0026] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, in which computer execution instructions are stored. When the computer execution instructions are executed by a processor, they are used to implement the image segmentation method based on the multi-task loss function provided in the embodiment of the first aspect above.
[0027] The technical solution proposed in the embodiment of the present invention optimizes the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to construct an image segmentation model; receives a target image acquisition instruction, obtains a target image to be segmented according to the target image acquisition instruction; and inputs the target image to be segmented into the image segmentation model for image segmentation. The technical solution proposed in the embodiment of the present invention solves the shortcomings of the existing image segmentation solutions, realizes the simultaneous optimization of the accuracy, connectivity and smoothness of the image segmentation model, significantly improves the performance of the image segmentation model, satisfies the refined positioning of the detected target, and improves the image segmentation effect. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] In order to more clearly illustrate and understand the technical solutions in the embodiments of the present invention, the following briefly introduces the background technology of the present invention and the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the contents of the embodiments of the present invention and these drawings without paying any creative work.
[0029] Figure 1A flow chart of an image segmentation method based on a multi-task loss function provided by an embodiment of the present invention;
[0030] Figure 2 A diagram showing the image segmentation effect of the solution of the present invention provided in an embodiment of the present invention;
[0031] Figure 3 A diagram showing the image segmentation effect of the solution of the present invention provided in an embodiment of the present invention;
[0032] Figure 4 A diagram showing the image segmentation effect of the solution of the present invention provided in an embodiment of the present invention;
[0033] Figure 5 A diagram showing the image segmentation effect of the solution of the present invention provided in an embodiment of the present invention;
[0034] Figure 6 A structural block diagram of an image segmentation device based on a multi-task loss function provided by an embodiment of the present invention; DETAILED DESCRIPTION
[0035] In order to make the technical problems solved by the present invention, the technical solutions adopted and the technical effects achieved clearer, the technical solutions of the embodiments of the present invention will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0036] Embodiment 1
[0037] Please refer to Figure 1 As stated, Figure 1 A flowchart of an image segmentation method based on a multi-task loss function provided by an embodiment of the present invention. As shown in the figure, the image segmentation method 100 based on a multi-task loss function in this embodiment includes:
[0038] S101. Optimizing the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to construct an image segmentation model;
[0039] S102. Receive a target image acquisition instruction, and obtain a target image to be segmented according to the target image acquisition instruction;
[0040] S103. Input the target image to be segmented into the image segmentation model for image segmentation.
[0041] Exemplarily, the multi-task loss function includes but is not limited to cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM.
[0042] Exemplarily, the method optimizes the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to construct an image segmentation model, specifically including:
[0043] Acquire an image to be segmented; use an image segmentation network to segment the image to be segmented to obtain a mask (2D or 3D) corresponding to the image to be segmented; optimize the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to construct an image segmentation model.
[0044] Exemplarily, the optimization of the accuracy, connectivity and smoothness of the image segmentation network by a multi-task loss function includes:
[0045] Calculate the cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM of the multi-task loss function; obtain the multi-task loss function according to the cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM.
[0046] Exemplarily, the optimization of the accuracy, connectivity and smoothness of the image segmentation network by a multi-task loss function includes:
[0047] The accuracy of the image segmentation model is optimized by the cross entropy loss CE; the Dice index of the image segmentation model is optimized by the Dice loss DC; the connectivity of the image segmentation model is optimized by the connectivity loss CC; and the smoothness of the image segmentation model is optimized by the smoothness loss SM.
[0048] Specifically, in actual application scenarios, the image segmentation network is a Unet (2D or 3D) network. The Unet network is called Unet because its network structure resembles the letter U. The left half is the feature extraction part, and the right half is the upsampling part, also known as the encoder-decoder structure (Encoder-Decoder). In this embodiment, the image to be segmented is input into the Unet network, and the corresponding feature map can be output by the previous layer of the last softmax layer of the Unet network. The value of the feature map is normalized to between 0 and 1 through the sigmoid function. It should be noted that the Sigmoid function, also known as the Logistic function, is an activation function commonly used in machine learning and neural networks.
[0049] The image segmentation method 100 based on the multi-task loss function proposed in the embodiment of the present invention optimizes the accuracy, connectivity and smoothness of the image segmentation network through the multi-task loss function to construct an image segmentation model; receives a target image acquisition instruction, obtains a target image to be segmented according to the target image acquisition instruction; and inputs the target image to be segmented into the image segmentation model for image segmentation. The technical solution proposed in this embodiment solves the shortcomings of the existing image segmentation solutions, realizes the simultaneous optimization of the accuracy, connectivity and smoothness of the image segmentation model, significantly improves the performance of the image segmentation model, satisfies the refined positioning of the detected target, and improves the image segmentation effect.
[0050] Embodiment 2
[0051] The image segmentation method based on the multi-task loss function in this embodiment specifically includes:
[0052] S201. Optimizing the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to construct an image segmentation model;
[0053] S202. Receive a target image acquisition instruction, and obtain a target image to be segmented according to the target image acquisition instruction;
[0054] S203. Input the target image to be segmented into the image segmentation model for image segmentation.
[0055] Exemplarily, the multi-task loss function includes but is not limited to cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM.
[0056] Exemplarily, the method optimizes the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to construct an image segmentation model, specifically including:
[0057] Acquire an image to be segmented; use an image segmentation network to segment the image to be segmented to obtain a mask (2D or 3D) corresponding to the image to be segmented; optimize the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to construct an image segmentation model.
[0058] Exemplarily, the optimization of the accuracy, connectivity and smoothness of the image segmentation network by a multi-task loss function includes:
[0059] Calculate the cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM of the multi-task loss function; obtain the multi-task loss function according to the cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM.
[0060] Exemplarily, the optimization of the accuracy, connectivity and smoothness of the image segmentation network by a multi-task loss function includes:
[0061] The accuracy of the image segmentation model is optimized by the cross entropy loss CE; the Dice index of the image segmentation model is optimized by the Dice loss DC; the connectivity of the image segmentation model is optimized by the connectivity loss CC; and the smoothness of the image segmentation model is optimized by the smoothness loss SM.
[0062] In this embodiment, for the multi-task loss function, the cross entropy loss CE is:
[0063]
[0064] Among them, $y_i$ is the true label, $p_i$ is the predicted probability, and $N$ is the number of samples.
[0065] Dice loss DC:
[0066]
[0067] Connectivity loss CC:
[0068]
[0069] Among them, $y_i$ is the true label, $p_i$ is the predicted probability, and $K$ is the connectivity window size
[0070] For the smoothing loss SM:
[0071] The smoothness loss includes two parts: basic smoothness and concavity detection.
[0072] Specifically, the basic smoothness loss is:
[0073]
[0074] Among them, (dx) and (dy) represent the absolute values of the gradient in the horizontal and vertical directions, respectively.
[0075] For sag detection loss:
[0076]
[0077] Where (N_{angles}) represents the number of ray sampling angles, (N_{steps}) represents the number of sampling points on each ray, and (r_{i,j}) represents the value of the (j)th sampling point on the (i)th ray.
[0078] The total loss is:
[0079] L=L basic +0.1×L concavity
[0080] In this embodiment, $\alpha$, $\beta$, and $\gamma$ are hyperparameters used to control the weights of different loss terms.
[0081] Specifically, in actual application scenarios, the image segmentation network is a Unet (2D or 3D) network. The Unet network is called Unet because its network structure resembles the letter U. The left half is the feature extraction part, and the right half is the upsampling part, also known as the encoder-decoder structure (Encoder-Decoder). In this embodiment, the image to be segmented is input into the Unet network, and the corresponding feature map can be output by the previous layer of the last softmax layer of the Unet network. The value of the feature map is normalized to between 0 and 1 through the sigmoid function. It should be noted that the Sigmoid function, also known as the Logistic function, is an activation function commonly used in machine learning and neural networks.
[0082] In the application of the present embodiment, experiments are conducted in the 3D ultrasound segmentation task of the prostate. Figure 2 As for Figure 5 The experimental results are shown. Obviously, the accuracy, connectivity and smoothness of the image segmentation model are optimized through the multi-task loss function, and the image segmentation achieves ideal results.
[0083] The image segmentation method based on a multi-task loss function proposed in this embodiment optimizes the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to construct an image segmentation model; receives a target image acquisition instruction, obtains a target image to be segmented according to the target image acquisition instruction; and inputs the target image to be segmented into the image segmentation model for image segmentation. The image segmentation method based on a multi-task loss function proposed in this embodiment solves the shortcomings of existing image segmentation schemes, realizes the simultaneous optimization of the accuracy, connectivity and smoothness of the image segmentation model, significantly improves the performance of the image segmentation model, satisfies the refined positioning of the detected target, and improves the image segmentation effect.
[0084] Embodiment 3
[0085] like Figure 6 As shown, Figure 6 The structure block diagram of the image segmentation device based on the multi-task loss function provided by the embodiment of the present invention. As shown in the figure, the image segmentation device 600 based on the multi-task loss function in this embodiment adopts the image segmentation method based on the multi-task loss function proposed in any one of the first embodiment or the second embodiment, and the device includes:
[0086] An image segmentation model construction module 601 is used to optimize the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to construct an image segmentation model;
[0087] The target image to be segmented acquisition module 602 is used to receive a target image acquisition instruction and acquire the target image to be segmented according to the target image acquisition instruction;
[0088] The image segmentation module 603 is used to input the target image to be segmented into the image segmentation model for image segmentation.
[0089] Exemplarily, the multi-task loss function includes but is not limited to cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM.
[0090] Exemplarily, the image segmentation model construction module 601 is specifically used to obtain an image to be segmented; perform image segmentation on the image to be segmented using an image segmentation network to obtain a mask (2D or 3D) corresponding to the image to be segmented; and optimize the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to construct an image segmentation model.
[0091] Exemplarily, the optimization of the accuracy, connectivity and smoothness of the image segmentation network by a multi-task loss function includes:
[0092] Calculate the cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM of the multi-task loss function; obtain the multi-task loss function according to the cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM.
[0093] Exemplarily, the optimization of the accuracy, connectivity and smoothness of the image segmentation network by a multi-task loss function includes:
[0094] The accuracy of the image segmentation model is optimized by the cross entropy loss CE; the Dice index of the image segmentation model is optimized by the Dice loss DC; the connectivity of the image segmentation model is optimized by the connectivity loss CC; and the smoothness of the image segmentation model is optimized by the smoothness loss SM.
[0095] In this embodiment, for the multi-task loss function, the cross entropy loss CE is:
[0096]
[0097] Among them, $y_i$ is the true label, $p_i$ is the predicted probability, and $N$ is the number of samples.
[0098] Dice loss DC:
[0099]
[0100] Connectivity loss CC:
[0101]
[0102] Among them, $y_i$ is the true label, $p_i$ is the predicted probability, and $K$ is the connectivity window size
[0103] For the smoothing loss SM:
[0104] The smoothness loss includes two parts: basic smoothness and concavity detection.
[0105] Specifically, the basic smoothness loss is:
[0106]
[0107] Among them, (dx) and (dy) represent the absolute values of the gradient in the horizontal and vertical directions, respectively.
[0108] For sag detection loss:
[0109]
[0110] Where (N_{angles}) represents the number of ray sampling angles, (N_{steps}) represents the number of sampling points on each ray, and (r_{i,j}) represents the value of the (j)th sampling point on the (i)th ray.
[0111] The total loss is:
[0112] L=L basic +0.1×L concavity
[0113] In this embodiment, $\alpha$, $\beta$, and $\gamma$ are hyperparameters used to control the weights of different loss terms.
[0114] Specifically, in actual application scenarios, the image segmentation network is a Unet (2D or 3D) network. The Unet network is called Unet because its network structure resembles the letter U. The left half is the feature extraction part, and the right half is the upsampling part, also known as the encoder-decoder structure (Encoder-Decoder). In this embodiment, the image to be segmented is input into the Unet network, and the corresponding feature map can be output by the previous layer of the last softmax layer of the Unet network. The value of the feature map is normalized to between 0 and 1 through the sigmoid function. It should be noted that the Sigmoid function, also known as the Logistic function, is an activation function commonly used in machine learning and neural networks.
[0115] The image segmentation device 600 based on a multi-task loss function proposed in this embodiment optimizes the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to construct an image segmentation model; receives a target image acquisition instruction, obtains a target image to be segmented according to the target image acquisition instruction; and inputs the target image to be segmented into the image segmentation model for image segmentation. The image segmentation device 600 based on a multi-task loss function proposed in this embodiment solves the shortcomings of existing image segmentation solutions, realizes simultaneous optimization of the accuracy, connectivity and smoothness of the image segmentation model, significantly improves the performance of the image segmentation model, satisfies the refined positioning of the detected target, and improves the image segmentation effect.
[0116] Embodiment 4
[0117] The present embodiment provides an electronic device, comprising a processor and a memory connected to the processor, wherein program data is stored in the memory, and the processor calls the program data stored in the memory to execute the image segmentation method based on a multi-task loss function as proposed in any one of the above-mentioned embodiments 1 or 2.
[0118] Embodiment 5
[0119] This embodiment provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are executed by a processor, they are used to implement the image segmentation method based on the multi-task loss function proposed in any of the above-mentioned embodiments 1 or 2.
[0120] It should be noted that the above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.
[0121] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims
1. An image segmentation method based on a multi-task loss function, characterized in that: include: The accuracy, connectivity and smoothness of the image segmentation network are optimized through a multi-task loss function to construct an image segmentation model. receiving a target image acquisition instruction, and acquiring a target image to be segmented according to the target image acquisition instruction; The target image to be segmented is input into the image segmentation model for image segmentation.
2. The image segmentation method based on multi-task loss function according to claim 1, characterized in that: The multi-task loss function includes but is not limited to cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM.
3. The image segmentation method based on multi-task loss function according to claim 1, characterized in that: The accuracy, connectivity and smoothness of the image segmentation network are optimized by the multi-task loss function to construct an image segmentation model, which specifically includes: Obtain the image to be segmented; Performing image segmentation on the image to be segmented using an image segmentation network to obtain a mask corresponding to the image to be segmented; The accuracy, connectivity and smoothness of the image segmentation network are optimized through a multi-task loss function to construct an image segmentation model.
4. The image segmentation method based on multi-task loss function according to claim 2, characterized in that: The accuracy, connectivity and smoothness of the image segmentation network are optimized by a multi-task loss function, including: Calculate the cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM of the multi-task loss function; obtain the multi-task loss function according to the cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM.
5. The image segmentation method based on multi-task loss function according to claim 4, characterized in that: The accuracy, connectivity and smoothness of the image segmentation network are optimized by a multi-task loss function, including: The accuracy of the image segmentation model is optimized by the cross entropy loss CE; the Dice index of the image segmentation model is optimized by the Dice loss DC; the connectivity of the image segmentation model is optimized by the connectivity loss CC; and the smoothness of the image segmentation model is optimized by the smoothness loss SM.
6. An image segmentation device based on a multi-task loss function, characterized in that: The device includes: An image segmentation model building module is used to optimize the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to build an image segmentation model; A target image acquisition module to be segmented, used for receiving a target image acquisition instruction and acquiring the target image to be segmented according to the target image acquisition instruction; The image segmentation module is used to input the target image to be segmented into the image segmentation model for image segmentation.
7. The image segmentation device based on multi-task loss function according to claim 6, characterized in that: The multi-task loss function includes but is not limited to cross entropy loss CE, Dice loss DC, connectivity loss CC, and smoothness loss SM.
8. The image segmentation device based on multi-task loss function according to any one of claims 6 or 7, characterized in that: The image segmentation model construction module is specifically used to obtain an image to be segmented; perform image segmentation on the image to be segmented using an image segmentation network to obtain a mask corresponding to the image to be segmented; and optimize the accuracy, connectivity and smoothness of the image segmentation network through a multi-task loss function to construct an image segmentation model.
9. An electronic device, characterized in that: The electronic device includes a processor and a memory connected to the processor, wherein program data is stored in the memory, and the processor calls the program data stored in the memory to execute the image segmentation method of the multi-task loss function as claimed in claim 1.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the image segmentation method of the multi-task loss function as claimed in claim 1.