An imaging method and apparatus based on an adversarial model

Abstract

The embodiment of the specification discloses an imaging method and device based on an adversarial model. The method comprises the following steps: obtaining reverse-time migration imaging of a profile to be imaged; inputting the reverse-time migration imaging into a pre-trained adversarial model to generate least-square reverse-time migration imaging of the profile to be imaged, wherein the adversarial model is pre-trained in the following manner: performing reverse-time migration imaging on a training profile to generate a first image, and constructing source domain data X comprising the first image; performing least-square reverse-time migration imaging on the training profile to generate a second image, and constructing target domain data Y comprising the second image; performing adversarial training on an initial model comprising a generator and a discriminator by using the source domain data and the target domain data to generate the pre-trained adversarial model; and thereby realizing that, for any reverse-time migration imaging, the pre-trained adversarial model can be used to convert the reverse-time migration imaging into high-precision least-square reverse-time migration imaging.

Description

TECHNICAL FIELD

[0001] The present specification relates to the field of oil and gas geophysical prospecting engineering, and in particular, to an imaging method and device based on an adversarial model. BACKGROUND

[0002] At present, the demand for exploration and development leads to an increase in the amount of seismic data acquisition and processing difficulty. The conventional migration imaging method is fast, but can only provide blurred imaging results and unreliable amplitude attributes. Although the least squares reverse-time migration imaging has higher accuracy, it requires a large amount of computing resources, and the efficiency is too low when the data amount is large.

[0003] Therefore, there is a need for a fast and high-precision imaging scheme. SUMMARY

[0004] The embodiments of the present specification provide an imaging method and device based on an adversarial model to solve the technical problem of needing a fast and high-precision imaging scheme.

[0005] To solve the above technical problems, one or more embodiments of the present specification are implemented as follows:

[0006] In a first aspect, the embodiments of the present specification provide an imaging method based on an adversarial model, comprising: obtaining reverse-time migration imaging for a profile to be imaged; inputting the reverse-time migration imaging into a pre-trained adversarial model to generate least squares reverse-time migration imaging for the profile to be imaged, wherein the adversarial model is pre-trained and generated based on the following manner:

[0007] Performing reverse-time migration imaging on a training profile to generate a first image, and constructing source domain data X including the first image; performing least squares reverse-time migration imaging on the training profile to generate a second image, and constructing target domain data Y including the second image; performing adversarial training on an initial model including a generator and a discriminator using the source domain data and the target domain data to generate the pre-trained adversarial model; wherein the generator is configured to obtain distribution characteristics of the target domain data, and generate a to-be-predicted image based on the distribution characteristics and the first image, and the discriminator is configured to determine whether the to-be-predicted image comes from the source domain data or the target domain data.

[0008] In a second aspect, one or more embodiments of the present specification provide an electronic device, comprising:

[0009] at least one processor; and

[0010] a memory connected in communication with the at least one processor; wherein

[0011] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to the first aspect.

[0012] The above at least one technical solution adopted by one or more embodiments of the present specification can achieve the following beneficial effects: by acquiring reverse-time migration imaging of a profile to be imaged; inputting the reverse-time migration imaging into a pre-trained adversarial model to generate least-square reverse-time migration imaging of the profile to be imaged, wherein the adversarial model is pre-trained based on the following manner: performing reverse-time migration imaging on a training profile to generate a first image, and constructing source domain data X including the first image; performing least-square reverse-time migration imaging on the training profile to generate a second image, and constructing target domain data Y including the second image; performing adversarial training on an initial model including a generator and a discriminator using the source domain data and the target domain data to generate the pre-trained adversarial model; wherein the generator is used to acquire distribution characteristics of the target domain data, and generate a to-be-predicted image based on the distribution characteristics and the first image, and the discriminator is used to judge whether the to-be-predicted image is from the source domain data or the target domain data, so as to realize that for any reverse-time migration imaging, it can be converted into high-precision least-square reverse-time migration imaging through the pre-trained adversarial model, realize fast high-precision imaging, and improve imaging effect. BRIEF DESCRIPTION OF DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present specification or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the present specification, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0014] Figure 1 a flow chart of an embodiment of the present application;

[0015] Figure 2 a network schematic diagram of an embodiment of the present application;

[0016] Figure 3 a generator structure diagram of an embodiment of the present application;

[0017] Figure 4 a discriminator structure diagram of an embodiment of the present application;

[0018] Figure 5 a conventional reverse-time migration imaging profile constituting a training set of the present application;

[0019] Figure 6a least-square reverse-time migration profile for a training set of the present application;

[0020] Figure 7 a velocity model for a testing set of the present application;

[0021] Figure 8 an imaging profile for a testing set of the present application;

[0022] Figure 9 a result comparison chart after imaging effect promotion for an embodiment of the present application. DETAILED DESCRIPTION

[0023] The embodiments of the present specification provide an imaging method and device based on an adversarial model.

[0024] In order for those skilled in the art to better understand the technical solutions in the present specification, the technical solutions in the embodiments of the present specification will be clearly and completely described below in conjunction with the drawings in the embodiments of the present specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present specification, all other embodiments obtained by those skilled in the art without creative labor should belong to the scope of protection of the present application.

[0025] As shown in Figure 1 , Figure 1 A flowchart of imaging based on an adversarial model provided by the embodiments of the present specification includes the following steps:

[0026] S101, acquiring reverse-time migration imaging for a profile to be imaged.

[0027] The profile to be imaged here can be any model or any observation data profile to be imaged, which can include any time point or spatial dimension, for example, any spatiotemporal profile can be taken from the reverse-time migration imaging of any observation data in actual production.

[0028] S103, inputting the reverse-time migration imaging into a pre-trained adversarial model to generate least-square reverse-time migration imaging for the profile to be imaged.

[0029] The adversarial model in the present application is pre-trained in the following manner: inverse time migration imaging is performed on a training profile to generate a first image, and source domain data X including the first image is constructed; least squares inverse time migration imaging is performed on the training profile to generate a second image, and target domain data Y including the second image is constructed; the initial model including a generator and a discriminator is adversarially trained using the source domain data and the target domain data to generate the pre-trained adversarial model; wherein the generator is configured to obtain distribution characteristics of the target domain data, and generate a to-be-predicted image based on the distribution characteristics and the first image, and the discriminator is configured to determine whether the to-be-predicted image is from the source domain data or the target domain data.

[0030] Specifically, the source domain data X and the target domain data Y can be constructed in advance using some models for imaging or already imaged underground images. For example, inverse time migration imaging and least squares inverse time migration imaging can be performed on the same observation data to obtain inverse time migration imaging results and least squares inverse time migration imaging results for the observation data. Then, imaging is performed on any profile from the inverse time migration imaging results and the least squares inverse time migration imaging results to obtain corresponding first images and second images, thereby constructing the source domain data X and the target domain data Y.

[0031] As shown in Figure 5 and Figure 6 , Figure 5 is an imaging profile (i.e., a first image) obtained by inverse time migration method for the model, Figure 6 is an imaging profile (i.e., a second image) corresponding to the same space-time obtained by least squares inverse time migration method for the model, thereby constructing a training set.

[0032] In the specific training process, a training set including a training data subset and a test data subset can be constructed. The training data subset and the test data subset both include source domain data and target domain data. The source domain data is composed of inverse time migration imaging profiles, and the imaging effect is general. The target domain data is composed of least squares inverse time migration imaging profiles, and the imaging effect is better.

[0033] In the embodiment of the present application, the initial model includes a one-way generative adversarial network GAN1 from source domain data to target domain data and a one-way GAN2 from target domain data to source domain data, and the GAN1 and the GAN2 share two generators G XY and G YX , the GAN1 includes a discriminator D X , and the GAN2 includes a discriminator D Y . As shown in Figure 2 , Figure 2A model network schematic diagram of an initial model of an embodiment of the present application.

[0034] In the adversarial training of the initial model containing the generator and the discriminator with the source domain data and the target domain data, the specific process includes:

[0035] S41, setting a learning rate and a training batch, inputting the source domain data and the target domain data into the initial model, and alternately training the generator and the discriminator;

[0036] S42, when the training batch is less than a threshold, updating the network parameters of the generator and the discriminator and continuing training; otherwise, jumping to S44;

[0037] S43, determining the gradient of the loss function of the generator and the discriminator, and passing the gradient back to the generator and the discriminator, and returning to step S42;

[0038] S44, training termination.

[0039] In this process, the GAN objective function of the one-way generative adversarial network GAN1 and GAN2 is: Where G is the generator, D is the discriminator, V(D, G) represents the value function about G and D, x is the input real data, which is subject to P data This distribution, z is random noise, is the distribution of random noise, E is the expectation, x~P data (x), z~P z (z) all represent the sampling of data subject to its distribution.

[0040] Specifically, the loss function of the generator is:

[0041] loss_G=loss_adv+λ cyc loss_cycle+λ id loss_identity, wherein loss_adv represents the adversarial loss function, loss_cycle represents the cycle consistency loss, loss_cycle represents the identity loss, λ cyc , λ id are the weights of the cycle consistency loss and the identity loss function respectively;

[0042] The loss function of the discriminator is:

[0043]

[0044] Wherein, loss_D1 and loss_D2 are two discriminators D X and D YThe loss function is shown in the following formula: loss = loss_real + loss_fake, loss_real is a real loss, which is the mean square error between the discrimination result of the discriminator for the real sample and True, loss_fake is a generated loss, which is the mean square error between the discrimination result of the discriminator for the generated sample and False. MSE represents the mean square error.

[0045] Further, the initial model shown in the formula is subjected to time, and the objective function thereof is: Figure 2

[0046]

[0047] The above formula is composed of two parts: one part loss GAN1 , loss GAN2 consists of an adversarial loss function, which has the same meaning as the ordinary generative adversarial network objective function, and tries to ensure that the data distribution generated by the generator is close to the target data. Another part is obtained by multiplying the cycle consistency loss function loss cycle and its weight λ, which ensures that the generated data from the source domain is mapped back by the second generation model, and the distribution is still the same.

[0048] The generator network is composed of an initialization convolution module, a down-sampling module, a residual network, an up-sampling module and an output layer (as shown in the formula). Figure 3 The initialization convolution block converts the input 3-channel 256x256 data matrix into the required number of layers for the convolution layer input in the down-sampling by filling and then performing convolution processing by 64 7x7 convolution kernels, and then normalizing and ReLU activation function. The down-sampling is completed by three convolution blocks, each of which includes a 2D convolution layer, an instance normalization layer and a ReLU activation function. The residual network is composed of 9 residual blocks, each of which includes 2 convolution layers, each of which is followed by an instance normalization layer and a ReLU activation function, and is connected through a residual connection. The up-sampling module includes three 2D transposed convolution layers, each of which is followed by an instance normalization layer and a ReLU activation function. The output layer is a 2D convolution layer using Tanh as the activation function, which generates an output of (256, 256, 3) parameters. The discriminator network is a convolutional neural network including five convolution blocks (as shown in the formula). Figure 4

[0049] In order to more specifically illustrate the method of the present application, the method of the present application is illustrated by taking the Marmousi model as an example. The Marmousi model contains a large number of reflecting interfaces and faults, and is commonly used for testing. The model velocity used is shown in the formula. Figure 7

[0050] ​​​After pre-training the adversarial model, we successfully improve the imaging profile of the conventional reverse time migration to the level of the imaging result of the least square reverse time migration method.

[0051] The migration imaging profile of the Marmousi model is shown in FIG. 1, where the left side is the imaging result of the conventional reverse time migration, and the right side is the imaging result of the least square reverse time migration method after 30 iterations. Figure 8 As can be seen from the figure, due to the influence of the illumination range of the seismic wave, the imaging result of the RTM is insufficient in the deep part and on both sides, and the reflection phase axis energy is weak; the imaging effect of the LSRTM is obviously improved, which is specifically manifested in that the imaging noise is suppressed, the imaging amplitude of the weak illumination area of the seismic wave is compensated, and the imaging resolution is improved. However, the least square reverse time migration still has many problems in practical application, and the huge amount of calculation and low convergence efficiency limit the application of the method in production.

[0052] As shown in FIG. 2, the left column in the figure is the imaging result of the conventional reverse time migration, the middle column is the imaging result of the least square reverse time migration, and the right column is the imaging result obtained by using the method based on the left reverse time migration from the imaging result and the pre-trained adversarial model. Figure 9

[0053] As can be seen from the comparison, compared with the imaging result of the conventional reverse time migration, the method used in the application can significantly improve the imaging effect itself and has better amplitude preservation. Compared with the imaging result of the least square reverse time migration, the application can quickly and in batches improve any number of reverse time migration imaging profiles through the well-trained adversarial network, and the network can be repeatedly used to improve the imaging effect of different imaging results. The imaging level is equivalent to that of the least square method, and therefore the efficiency is better.

[0054] In a second aspect, the embodiments of the present specification also provide an electronic device, including: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method as described in the first aspect.

[0055] ​In the 1990s, it was quite obvious to distinguish whether an improvement in a technology was in hardware (e.g., improvement in circuit structures of diodes, transistors, switches, etc.) or in software (improvement in method flow). However, as technology has evolved, many improvements in method flow today can be considered as direct improvements in hardware circuit structures. Designers almost always obtain the corresponding hardware circuit structures by programming the improved method flow into hardware circuits. Therefore, it cannot be said that an improvement in a method flow cannot be implemented by hardware entity modules. For example, a programmable logic device (PLD) (e.g., a field programmable gate array (FPGA)) is an integrated circuit whose logic function is determined by user programming of the device. A digital system is "integrated" on a PLD by the designer programming it, rather than by asking a chip manufacturer to design and fabricate a custom integrated circuit chip. Moreover, instead of manually fabricating integrated circuit chips, this programming is now mostly implemented by "logic compiler" software, which is similar to software compilers used in program development, and the original code to be compiled is written in a specific programming language, which is called a hardware description language (HDL), and there are many such languages, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc., and the most commonly used are VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should be aware that, as long as the method flow is logically programmed in the above-mentioned hardware description languages and programmed into an integrated circuit, a hardware circuit implementing the logical method flow can be easily obtained.

[0056] The controller can be implemented in any suitable way, e.g. the controller can take the form of a microprocessor or processor and a computer readable medium storing computer readable program code, e.g. software or firmware, executable by the (micro)processor, logic gates, switches, an application specific integrated circuit (ASIC), a programmable logic controller and an embedded microcontroller, examples of controllers include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91 SAM, Microchip PIC18F26K20 and Silicone Labs C8051F320, the memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also know that in addition to being implemented in pure computer readable program code form, the controller can perfectly well be implemented by means of logic programmed into logic gates, switches, application specific integrated circuits, programmable logic controllers and embedded microcontrollers, etc. to perform the same functions. The controller can thus be considered as a hardware component, and the means comprised therein for performing various functions can be considered as structures within the hardware component. Alternatively, or even additionally, the means for performing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0057] The systems, apparatuses, modules or units illustrated by the above embodiments can be implemented by computer chips or entities, or products with certain functions. A typical implementation device is a computer. Specifically, the computer can be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0058] For the sake of description, the above apparatuses are described in various units by functions respectively. Of course, the functions of the units can be implemented in one or more software and / or hardware in the implementation of the present specification.

[0059] Those skilled in the art will understand that the embodiments of the present specification can be provided as a method, a system or a computer program product. Therefore, the embodiments of the present specification can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present specification can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0060] The specification is presented with reference to flow diagrams and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the specification. It will be understood that each block of the flow diagrams and / or block diagrams, and combinations of blocks in the flow diagrams and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing element or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flow diagrams and / or block diagrams block or blocks. Figure 1 The flow diagrams and / or block diagrams in the specification can present a method, apparatus or computer program product according to embodiments of the specification. Flow diagrams and / or block diagrams can also present a method, apparatus or computer program product to achieve functions specified in flow diagrams and / or block diagrams block or blocks. Figure 1 The flow diagrams and / or block diagrams in the specification can present a method, apparatus or computer program product according to embodiments of the specification. Flow diagrams and / or block diagrams can also present a method, apparatus or computer program product to achieve functions specified in flow diagrams and / or block diagrams block or blocks.

[0061] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow diagrams and / or block diagrams block or blocks. Figure 1 The flow diagrams and / or block diagrams in the specification can present a method, apparatus or computer program product according to embodiments of the specification. Flow diagrams and / or block diagrams can also present a method, apparatus or computer program product to achieve functions specified in flow diagrams and / or block diagrams block or blocks. Figure 1 The flow diagrams and / or block diagrams in the specification can present a method, apparatus or computer program product according to embodiments of the specification. Flow diagrams and / or block diagrams can also present a method, apparatus or computer program product to achieve functions specified in flow diagrams and / or block diagrams block or blocks.

[0062] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow diagrams and / or block diagrams block or blocks. Figure 1 The flow diagrams and / or block diagrams in the specification can present a method, apparatus or computer program product according to embodiments of the specification. Flow diagrams and / or block diagrams can also present a method, apparatus or computer program product to achieve functions specified in flow diagrams and / or block diagrams block or blocks. Figure 1 The flow diagrams and / or block diagrams in the specification can present a method, apparatus or computer program product according to embodiments of the specification. Flow diagrams and / or block diagrams can also present a method, apparatus or computer program product to achieve functions specified in flow diagrams and / or block diagrams block or blocks.

[0063] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0064] The memory can include non-persistent memory and / or storage mechanisms such as, for example, random access memory (RAM), non-volatile memory (NVM), and / or a persistent memory such as, for example, read-only memory (ROM) or flash memory. The memory is an example of computer-readable media.

[0065] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0066] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or apparatus that comprises a list of elements does not only include those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0067] The specification can be described in the general context of computer-executable instructions, such as program modules, executed by computers. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The specification can also be practiced in a distributed computing environment, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0068] Each embodiment in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other, and each embodiment focuses on the difference from other embodiments. In particular, for device, equipment, non-volatile computer storage medium embodiments, because they are basically similar to method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiment.

[0069] The above-described embodiments of the application have special structure and can achieve the desired results. Other embodiments can have different structures and achieve the same results. The purpose of the above-described embodiments is to illustrate the principles of the application and not to limit the scope of the application. The scope of the application is defined by the claims and their equivalents. Other embodiments are within the scope of the claims.

[0070] The above description is merely illustrative of the embodiments of the present application and is not intended to limit the scope of the present application. Various modifications can be made by those skilled in the art based upon the teachings disclosed herein. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of the present application shall fall within the scope of the claims of the present application.

Claims

1. An imaging method based on an adversarial model, comprising: obtaining reverse-time migration imaging for a profile to be imaged; inputting the reverse-time migration imaging into a pre-trained adversarial model to generate least-square reverse-time migration imaging for the profile to be imaged, wherein the adversarial model is pre-trained based on the following manner: performing reverse-time migration imaging on a training profile to generate a first image, and constructing source domain data X comprising the first image; performing least-square reverse-time migration imaging on the training profile to generate a second image, and constructing target domain data Y comprising the second image; performing adversarial training on an initial model comprising a generator and a discriminator using the source domain data and the target domain data to generate the pre-trained adversarial model, wherein the generator is configured to obtain distribution characteristics of the target domain data, and generate a predicted image based on the distribution characteristics and the first image, and the discriminator is configured to determine whether the predicted image is from the source domain data or the target domain data; The initial model comprising the generator and the discriminator includes a one-way generative adversarial network (GAN1) from source domain data to target domain data and a one-way generative adversarial network (GAN2) from target domain data to source domain data, the GAN1 and the GAN2 share two generators G XY and G YX , and the GAN1 includes a discriminator D X , and the GAN2 includes a discriminator D Y ; a loss function of the generator is: loss_G = loss_adv + λ cyc loss_cycle + λ id loss_identity, wherein loss_adv represents an adversarial loss function, loss_cycle represents a cycle consistency loss, loss_identity represents an identity loss, λ cyc , and λ id are weights of the cycle consistency loss and the identity loss function, respectively. a loss function of the discriminator is: Wherein, loss_D1, loss_D2 are loss functions of two discriminators D X and D Y , loss_real is a real loss, which is the mean square error of the discrimination result of the discriminator on the real sample and True, loss_fake is a generated loss, which is the mean square error of the discrimination result of the discriminator on the output sample of the generator and False, and MSE represents the mean square error.

2. The method of claim 1, wherein, performing adversarial training on the initial model comprising the generator and the discriminator using the source domain data and the target domain data, comprising: S41, setting a learning rate and a training batch, inputting the source domain data and the target domain data into the initial model, and alternately training the generator and the discriminator; S42, when the training batch is less than a threshold, updating network parameters of the generator and the discriminator and continuing training; otherwise, jumping to S44; S43, determining gradients of the loss functions of the generator and the discriminator, and passing the gradients back to the generator and the discriminator, and returning to step S42; S44, terminating training.

3. The method of claim 1, wherein, The GAN objective function of the one-way generative adversarial network GAN1 and GAN2 is: where G is a generator, D is a discriminator, V(D, G) represents a value function with respect to G and D, x is input real data, and it is subject to P data This distribution, z is random noise, is a distribution of random noise, E is expectation, x ~ P data (x), z ~ P z (x), z ~ P 4. The method of claim 3, wherein an objective function of the initial model is: wherein loss GAN1 is an adversarial loss value generated by GAN1 based on the GAN objective function, the loss GAN2 is an adversarial loss value generated by GAN2 based on the GAN objective function, loss cycle is a cycle consistency loss value, wherein, |||1 represents an L1 norm.

5. An electronic device, comprising: at least one processor; and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 4.

Images