An ai-based content generation method and generation system

Abstract

The application relates to an AI-based content generation method and system, which comprises the following steps: acquiring a separated image and a background image, and separating a feature object from the separated image; establishing a first light model belonging to the separated image and a second light model belonging to the background image; restoring the feature object by using the first light model to obtain a basic feature object; processing the basic feature object by using the second light model to obtain a fused feature object; transferring the fused feature object to the background image, dividing a fusion domain at a surrounding edge of the fused feature object, and processing the fusion domain by using a generative adversarial network to make the fused feature object fuse with the background image. The AI-based content generation method and system disclosed by the application use an environmental feature combined with a generative adversarial network to convert and generate content, and the method can limit the generative adversarial network to obtain high-true-content.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to an AI-based content generation method and system. BACKGROUND

[0002] The research field of AI (Artificial Intelligence) is extensive and in-depth, covering multiple subfields such as machine learning, deep learning, natural language processing, computer vision, robotics, etc., and these technologies have been widely applied in our daily life, such as voice recognition in smartphones, self-driving cars, smart home devices, and medical diagnosis systems, etc.

[0003] In the field of productivity, AI is also increasingly widely used, for example, in the field of content (text, pictures, videos, etc.) generation, AI can widely replace repetitive operations and generate the required content according to instructions, but in the generation of picture and video content, AI has the problem of generating unreal content.

[0004] Taking the conversion generation as an example, there is a big difference between the converted content and the converted content, and the simple copy and paste method will produce unreal, at this time, the generative adversarial network is introduced to correct, which can solve the problem to a certain extent, but it is easy to appear local optimal solution problem. SUMMARY

[0005] The present application provides an AI-based content generation method and system, which uses environmental features combined with a generative adversarial network to convert content, which can limit the generative adversarial network to obtain high-fidelity content.

[0006] The above-mentioned purpose of the present application is realized by the following technical scheme:

[0007] In a first aspect, the present application provides an AI-based content generation method, comprising:

[0008] Obtaining a separated image and a background image and separating a feature object from the separated image;

[0009] Establishing a first light model belonging to the separated image and a second light model belonging to the background image;

[0010] Restoring the feature object using the first light model to obtain a basic feature object;

[0011] Processing the basic feature object using the second light model to obtain a fusion feature object;

[0012] Transferring the fusion feature object to the background image and dividing the fusion domain at the surrounding edge of the fusion feature object;

[0013] The fusion domain is processed using a generative adversarial network to make the fusion feature object and the background image fused.

[0014] In a possible implementation manner of the first aspect, the establishing the first illumination model belonging to the separated image comprises:

[0015] After the feature object is processed in grayscale, a plurality of regions are selected on the feature object, and each region includes at least one pixel point;

[0016] A reference line is established using the regions;

[0017] The reference line is rotated, and a pixel value change trend on the reference line is calculated, so that at least two reference lines with the same pixel value change trend are parallel or intersected;

[0018] The first illumination model is established using the rotated reference line.

[0019] In a possible implementation manner of the first aspect, the restoring the feature object using the first illumination model to obtain the basic feature object comprises:

[0020] The first illumination model is adjusted, so that light sources in the first illumination model tend to be consistent, and a restoration direction line is established for the adjusted first illumination model;

[0021] An enhancement intensity or a reduction intensity on the restoration direction line is calculated;

[0022] The feature object is adjusted using the enhancement intensity or the reduction intensity to obtain the basic feature object.

[0023] In a possible implementation manner of the first aspect, after the enhancement intensity or the reduction intensity on the restoration direction line is calculated, the method further comprises:

[0024] An intersection influence area and a blank influence area of the restoration direction line are determined;

[0025] A restoration direction line associated with the blank influence area is determined;

[0026] A secondary restoration direction line is created using the associated restoration direction line, and the blank influence area is adjusted based on the secondary restoration direction line.

[0027] In a possible implementation manner of the first aspect, the processing the basic feature object using the second illumination model to obtain the fusion feature object comprises:

[0028] A brightness parameter of a transfer position of the fusion feature object transferred onto the background image is determined;

[0029] The basic feature object is adjusted using the brightness parameter;

[0030] adjusting the base feature object again using the second illumination model;

[0031] obtaining the adjusted feature object, denoted as a fusion feature object.

[0032] In a possible implementation manner of the first aspect, determining the luminance parameter of the fusion feature object transferred to the transfer position on the background image comprises:

[0033] sequentially establishing a plurality of annular analysis reference lines inside the feature object and transferring the plurality of annular analysis reference lines to the background image in a direction close to the edge of the feature object, with a previous annular analysis reference line inside a next annular analysis reference line;

[0034] calculating a pixel value change on the annular analysis reference line and determining the brightest point and the dimmest point according to the pixel value change;

[0035] creating the luminance parameter using the brightest point and the dimmest point.

[0036] In a possible implementation manner of the first aspect, when the fusion domain is processed using the generative adversarial network, the method further comprises adjusting the width of the fusion domain.

[0037] In a second aspect, the present application provides an AI-based content generation device, comprising:

[0038] a first processing unit configured to acquire a separation image and a background image and separate a feature object from the separation image;

[0039] an illumination model establishing unit configured to establish a first illumination model belonging to the separation image and a second illumination model belonging to the background image;

[0040] a second processing unit configured to restore the feature object using the first illumination model to obtain a base feature object;

[0041] a third processing unit configured to process the base feature object using the second illumination model to obtain a fusion feature object;

[0042] a fourth processing unit configured to transfer the fusion feature object to the background image and divide a fusion domain at a surrounding edge of the fusion feature object;

[0043] a fusion processing unit configured to process the fusion domain using a generative adversarial network to fuse the fusion feature object with the background image.

[0044] In a third aspect, the present application provides a content generation system, comprising:

[0045] one or more memories configured to store instructions; and

[0046] one or more processors configured to invoke and run the instructions from the memory to perform the method as claimed in the first aspect and any possible implementation of the first aspect.

[0047] In a fourth aspect, a computer readable storage medium is provided, the computer readable storage medium comprising:

[0048] a program which, when executed by a processor, causes the method as claimed in the first aspect and any possible implementation of the first aspect to be performed.

[0049] In a fifth aspect, a computer program product is provided, comprising program instructions which, when executed by a computing device, cause the method as claimed in the first aspect and any possible implementation of the first aspect to be performed.

[0050] In a sixth aspect, a chip system is provided, the chip system comprising a processor configured to implement the functions of the above aspects, for example, generating, receiving, sending, or processing data and / or information involved in the above methods.

[0051] The chip system can be composed of a chip, or can include a chip and other discrete devices.

[0052] In a possible design, the chip system further includes a memory, the memory configured to store necessary program instructions and data. The processor and the memory can be decoupled and disposed on different devices, connected through a wired or wireless manner, or the processor and the memory can be coupled on the same device. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 is a schematic block diagram of a step flow of a content generation method provided in the present application.

[0054] Figure 2 is a schematic diagram of establishing a region provided in the present application.

[0055] Figure 3 is a schematic diagram of rotating a reference line provided in the present application.

[0056] Figure 4 is a schematic diagram of obtaining a point light source and a range light source provided in the present application.

[0057] Figure 5 is a schematic diagram of a blank influence region provided in the present application.

[0058] Figure 6 is a schematic diagram of establishing a secondary restoration direction line based on an edge of a coverage region provided in the present application.

[0059] Figure 7 is a schematic diagram provided by the present application for determining the brightest point and the weakest point of illumination according to the change of pixel value on the annular analysis reference line. DETAILED DESCRIPTION

[0060] First, the local optimal solution of the generative adversarial network is explained.

[0061] GANs are composed of two mutually opposed neural networks: the generator (Generator) and the discriminator (Discriminator). The goal of the generator is to generate fake data that approximates real data, while the goal of the discriminator is to distinguish between real data and fake data.

[0062] This generation of adversarial training method enables GANs to learn the distribution of data and generate high-quality new data. However, since the generator and the discriminator need to play multiple rounds of adversarial games during training, it is easy to fall into a local optimal solution.

[0063] The specific reasons are as follows:

[0064] Complex training process: The training process of GANs involves the mutual competition and adaptation of two networks, and this complex dynamic relationship may lead to training falling into a local optimum.

[0065] Loss function design: The loss function of GANs is usually designed based on maximizing the error of the discriminator and minimizing the error of the generator. However, this design may lead to instability in the training process and the emergence of local optimal solutions.

[0066] Model structure selection: Different model structures have a great influence on the training effect of GANs. If the model structure is not properly designed, it may also lead to training falling into a local optimum.

[0067] The technical solutions in the present application are further described in detail in combination with the drawings.

[0068] The present application discloses an AI-based content generation method, please refer to Figure 1 In some examples, the AI-based content generation method disclosed in the present application includes the following steps:

[0069] S101, obtaining a separation image and a background image and separating a feature object from the separation image;

[0070] S102, establishing a first illumination model belonging to the separation image and a second illumination model belonging to the background image;

[0071] S103, restoring the feature object using the first illumination model to obtain a basic feature object;

[0072] S104, processing the basic feature object using the second light model to obtain a fusion feature object;

[0073] S105, transferring the fusion feature object to the background image and dividing a fusion domain at a surrounding edge of the fusion feature object;

[0074] S106, processing the fusion domain using a generative adversarial network to make the fusion feature object fuse with the background image.

[0075] Overall, the technical solution disclosed in the present application first processes the feature object to be transferred using a light model, and then processes the fusion domain using a generative adversarial network, rather than directly processing the entire feature object using the generative adversarial network.

[0076] The advantage of this processing method is that it greatly reduces the processing area, so that the generative adversarial network only needs to process a limited fusion domain.

[0077] In step S101, first, the separated image and the background image are obtained, and the feature object is separated from the separated image, and then in step S102, the first light model belonging to the separated image and the second light model belonging to the background image are established.

[0078] In step S103, the feature object is restored using the first light model to obtain a basic feature object. The purpose of restoring the feature object is to remove the influence of light in the environment to which the separated image belongs on the basic feature object.

[0079] In step S104, the second light model is used to process the basic feature object to obtain a fusion feature object. The purpose of processing the basic feature object using the second light model is to apply the light in the environment to which the background image belongs to the basic feature object.

[0080] In step S105, the fusion feature object is transferred to the background image and a fusion domain is divided at the surrounding edge of the fusion feature object. The fusion domain needs to be processed using a generative adversarial network, that is, the content in step S106. After the processing of the fusion domain by the generative adversarial network, the fusion feature object can be fused with the background image.

[0081] As can be seen from the above description, the processing of the feature object using the first light model and the second light model is based on the background image to perform a fusion processing on the feature object; and the processing of the fusion domain using the generative adversarial network is based on the background image to perform a secondary fusion processing on the feature object.

[0082] The purpose of the two fusion processes is to solve the overall problem and the local problem in the fusion process, respectively.

[0083] Before step S101, if the resolution of the separated image and the background image are different, the resolution needs to be adjusted to be the same first, and then the subsequent steps are performed. When adjusting the resolution, the resolution of the separated image is reduced when the resolution of the background image is low, and the resolution of the background image is reduced when the resolution of the separated image is low.

[0084] In some examples, the specific way of establishing the first illumination model belonging to the separated image is:

[0085] S201, selecting a plurality of regions on the feature object after gray processing of the feature object, each region including at least one pixel point;

[0086] S202, establishing a reference line using the regions;

[0087] S203, rotating the reference line and calculating the pixel value change trend on the reference line, so that at least two reference lines with the same pixel value change trend are parallel or intersected;

[0088] S204, establishing the first illumination model using the rotated reference line.

[0089] In steps S201 and S202, the feature object is first gray processed, and the purpose of the gray processing is to reduce the data amount of the feature object, and then a plurality of regions are selected on the feature object. Please refer to Figure 2 The selected regions are as close to the edges of the feature object as possible and are randomly distributed.

[0090] Each region includes at least one pixel point, and the region is used to establish a reference line, and the reference line takes the region as a starting point or passes through the starting point. When the number of pixel points included in the region is multiple, the center position point of the region is used to establish the reference line.

[0091] In step S203, please refer to Figure 3 The reference line will be rotated and the pixel value change trend on the reference line will be calculated, so that at least two reference lines with the same pixel value change trend are parallel or intersected. The pixel value change trend here has two cases of increase and decrease.

[0092] For the change trend of the pixel value change, one calculation method is to randomly select a point or a segment (the values of the pixel points in the segment are processed by mean value) on the reference line, and then calculate the change trend of the pixel value change according to the order.

[0093] Of course, the slope corresponding to the change trend obtained at this time will not be a stable value, and in this application, the value is required to fluctuate within a certain range, and the general allowed error is between ±0.2-0.6. Of course, the error can also be changed, which is only an example here and is not limited.

[0094] The purpose of rotating the reference line is to find the light source affecting the feature object, the number of light sources may be one or more, and the types of light sources also exist in two, which are bright light source and dark light source. The role of bright light source is to emit light, and the role of dark light source is to absorb light.

[0095] The obtained values of bright light source and dark light source are directly used as the values at the corresponding positions on the reference line.

[0096] In addition, the result of rotation is to make the trend of pixel value change (increase and decrease) on the reference line tend to be stable, that is, the trend of pixel value change on the reference line at this time is linear change.

[0097] After rotation, the reference line will appear in two cases, which are intersection and parallel, intersection represents point light source, and parallel represents range light source.

[0098] Finally, the first light model is established using the rotated reference line. The first light model is based on the foregoing record to record the point light source and the range light source, and at the same time record the intensity of the point light source and the range light source.

[0099] The specific way to restore the feature object using the first light model is:

[0100] S301, adjusting the first light model, making the light sources in the first light model consistent and establishing a restoration direction line for the adjusted first light model;

[0101] S302, calculating the intensity or reduction degree on the restoration direction line;

[0102] S303, adjusting the feature object using the intensity or reduction degree to obtain the basic feature object.

[0103] Specifically, first, the first light model needs to be adjusted, and then the light sources in the first light model are made consistent and a restoration direction line is established for the adjusted first light model. Here, making the light sources in the first light model consistent means making the intensity of the light sources in the first light model consistent, and of course the distance parameter needs to be considered here.

[0104] Then, the intensity or reduction degree on the restoration direction line is calculated. The calculation method here is obtained according to the linear calculation method. Finally, the feature object is adjusted using the intensity or reduction degree. The specific adjustment process is to make the intensity of the light sources in the first light model consistent. The value here uses the brightness parameter of the fusion feature object transferred to the transfer position on the background image.

[0105] The enhancement corresponds to a dark light source, and the reduction corresponds to a bright light source. The enhancement and the reduction are a constant value, that is, the slope of the diagonal line. The influence of the light source on the feature object stops when the value on the restoration direction line is reduced to zero or when the edge of the feature object is crossed.

[0106] After the enhancement or the reduction on the restoration direction line is calculated, the following is added:

[0107] S401, determine the intersection influence area and the blank influence area of the restoration direction line;

[0108] S402, determine the restoration direction line associated with the blank influence area;

[0109] S403, create a secondary restoration direction line using the associated restoration direction line and adjust the blank influence area based on the secondary restoration direction line.

[0110] In steps S401 to S403, the intersection influence area and the blank influence area of the restoration direction line are first determined. The intersection influence area refers to an area affected by at least two light sources, and the blank influence area refers to an area not affected by any light source.

[0111] For the intersection influence area, the associated light source is used for processing, and for the blank influence area, a secondary restoration direction line is created using the associated restoration direction line and the blank influence area is adjusted based on the secondary restoration direction line.

[0112] Please refer to Figure 5 and Figure 6 The specific process is to establish a secondary restoration direction line based on the edge of the intersection influence area. The secondary restoration direction line is perpendicular to the associated restoration direction line (the dashed line in Figure 6 ), and the starting parameter of the secondary restoration direction line is the parameter at the position of the starting point on the restoration direction line.

[0113] In some examples, the specific steps of processing the base feature object using the second lighting model to obtain the fused feature object are as follows:

[0114] S501, determine the luminance parameter of the fused feature object transferred to the transfer position on the background image;

[0115] S502, adjust the base feature object using the luminance parameter;

[0116] S503, adjust the base feature object again using the second lighting model;

[0117] S504, obtain the adjusted feature object, and the adjusted feature object is denoted as the fused feature object.

[0118] In steps S501 to S504, the base feature object is adjusted twice using the brightness parameter and the second illumination model respectively, the first adjustment is to adjust the whole of the base feature object, and the second adjustment is to adjust the local part of the base feature object.

[0119] Through the above-mentioned adjustment, the influence of the illumination on the base feature object on the separated image can be eliminated while the base feature object is affected by the illumination on the background image.

[0120] The specific way of determining the brightness parameter of the fusion feature object transferred to the transfer position on the background image is:

[0121] S601, sequentially establishing a plurality of annular analysis reference lines inside the feature object and transferring the plurality of annular analysis reference lines to the background image, in the direction close to the edge of the feature object, the previous annular analysis reference line is inside the next annular analysis reference line;

[0122] S602, calculating the pixel value change on the annular analysis reference line and determining the brightest point and the dimmest point according to the pixel value change;

[0123] S603, creating the brightness parameter using the brightest point and the dimmest point.

[0124] Please refer to Figure 7 In steps S601 to S603, first, a plurality of annular analysis reference lines (dashed lines in Figure 7 ) are sequentially established inside the feature object and transferred to the background image, and then the sequentially established plurality of annular analysis reference lines are used to obtain the brightness parameter.

[0125] The specific way is to calculate the pixel value change on the annular analysis reference line and determine the brightest point and the dimmest point according to the pixel value change, and then create the brightness parameter using the brightest point and the dimmest point, the value of the brightest point spreads to the surrounding according to a linear relationship, the value in the spreading process tends to decrease, and the value of the dimmest point also spreads to the surrounding according to a linear relationship, the value in the spreading process tends to increase.

[0126] When the two meet, the value decrease speed and the value increase speed in the spreading process are adjusted in reverse after mean value processing at the meeting position.

[0127] In some examples, when the fusion domain is processed using the generative adversarial network, the width of the fusion domain is also adjusted, and the adjustment of the width of the fusion domain occurs when the processing result of the generative network cannot be recognized by the adversarial network.

[0128] This is because the width of the fusion domain directly limits the processing range of the generation network, and too narrow width will make the change trend steep in the fusion domain, so it is necessary to appropriately increase the width of the fusion domain, and vice versa, when the processing result of the generation network is recognized by the adversarial network, the width of the fusion domain needs to be reduced.

[0129] The application also provides an AI-based content generation device, comprising:

[0130] A first processing unit is configured to acquire a separated image and a background image and separate a feature object from the separated image;

[0131] A light model establishing unit is configured to establish a first light model belonging to the separated image and a second light model belonging to the background image;

[0132] A second processing unit is configured to restore the feature object using the first light model to obtain a basic feature object;

[0133] A third processing unit is configured to process the basic feature object using the second light model to obtain a fused feature object;

[0134] A fourth processing unit is configured to transfer the fused feature object to the background image and divide a fusion domain at the surrounding edge of the fused feature object;

[0135] A fusion processing unit is configured to process the fusion domain using a generative adversarial network to fuse the fused feature object with the background image.

[0136] Further, it further comprises:

[0137] A region selecting unit is configured to select a plurality of regions on the feature object after performing gray processing on the feature object, and each region comprises at least one pixel point;

[0138] A reference line establishing unit is configured to establish a reference line using the regions;

[0139] A reference line rotating unit is configured to rotate the reference line and calculate the pixel value change trend on the reference line, so that at least two reference lines with the same pixel value change trend are parallel or intersected;

[0140] A first light model establishing unit is configured to establish the first light model using the rotated reference line.

[0141] Further, it further comprises:

[0142] A first light model adjusting unit is configured to adjust the first light model, so that the light sources in the first light model tend to be consistent, and a restored direction line is given to the adjusted first light model;

[0143] The first computing unit is configured to calculate the enhancement or reduction degree on the restoration direction line.

[0144] The first adjusting unit is configured to adjust the feature object using the enhancement or reduction degree to obtain a basic feature object.

[0145] Further, the method further comprises:

[0146] The first region determining unit is configured to determine the cross-influence region and the blank-influence region of the restoration direction line.

[0147] The second region determining unit is configured to determine the restoration direction line associated with the blank-influence region.

[0148] The second adjusting unit is configured to create a secondary restoration direction line using the associated restoration direction line and adjust the blank-influence region based on the secondary restoration direction line.

[0149] Further, the method further comprises:

[0150] The parameter determining unit is configured to determine a brightness parameter of a transfer position of the fusion feature object transferred onto the background image.

[0151] The third adjusting unit is configured to adjust the basic feature object using the brightness parameter.

[0152] The fourth adjusting unit is configured to adjust the basic feature object again using the second illumination model.

[0153] The obtaining unit is configured to obtain the adjusted feature object, and the adjusted feature object is referred to as a fusion feature object.

[0154] Further, the method further comprises:

[0155] The first processing unit is configured to sequentially establish a plurality of annular analysis reference lines inside the feature object and transfer the plurality of annular analysis reference lines onto the background image in a direction close to the edge of the feature object, wherein a previous annular analysis reference line is located inside a next annular analysis reference line.

[0156] The second processing unit is configured to calculate a pixel value change on the annular analysis reference line and determine a brightest point and a dimmest point according to the pixel value change.

[0157] The brightness parameter creating unit is configured to create the brightness parameter using the brightest point and the dimmest point.

[0158] Further, when the fusion domain is processed using the generative adversarial network, the width of the fusion domain is adjusted.

[0159] In an example, the units in any of the above apparatuses can be one or more integrated circuits, configured to implement one or more of the above methods, e.g., one or more application specific integrated circuits (ASICs), or, one or more digital signal processors (DSPs), or, one or more field programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.

[0160] Further, when the units in the apparatuses can be implemented in the form of a processing element scheduler, the processing element can be a general purpose processor, such as a central processing unit (CPU) or other processor that can invoke a program. Further, these units can be integrated together, implemented in the form of a system-on-a-chip (SOC).

[0161] In the present application, various messages / information / devices / network elements / systems / apparatuses / actions / operations / processes / concepts, etc. can be named, and it can be understood that these specific names do not constitute a limitation on the related objects, and the names can be changed according to the scene, context or use habit, etc. The technical meaning of the technical terms in the present application should be mainly determined according to the function and technical effect embodied / implemented in the technical scheme.

[0162] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, apparatus and unit can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0163] In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method can be implemented in other ways. For example, the above-described apparatus embodiments are merely schematic, and the division of the units is merely a logical function division. In actual implementation, another division manner can be used, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0164] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Part or all of the units may be selected according to actual needs to achieve the purpose of the embodiments.

[0165] Those skilled in the art can appreciate that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solutions. A skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0166] It should also be understood that in various embodiments of the present application, first, second, etc. are only to represent that a plurality of objects are different. For example, the first time window and the second time window are only to represent different time windows. The above first, second, etc. should not have any impact on the time window itself, and should not limit the embodiments of the present application.

[0167] It should also be understood that in various embodiments of the present application, the terms and / or descriptions of different embodiments are consistent and can be referred to each other if there is no special description and logical conflict, and the technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0168] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the essential part of the prior art or the part of the technical solutions can be embodied in the form of software products, which are stored in a computer readable storage medium, including a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in various embodiments of the present application. The foregoing computer readable storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or optical disk and various program code storage media.

[0169] The present application also provides a content generation system, the system comprising:

[0170] one or more memories for storing instructions; and

[0171] one or more processors for invoking and running the instructions from the memory to perform the methods described above.

[0172] The present application also provides a computer program product including instructions which, when executed by a terminal device and a network device, cause the terminal device and the network device to perform the operations of the terminal device and the network device corresponding to the above methods.

[0173] The present application also provides a chip system including a processor for implementing the functions involved in the above description, such as generating, receiving, sending, or processing the data and / or information involved in the above methods.

[0174] The chip system can be composed of a chip, or can include a chip and other discrete devices.

[0175] The processor mentioned in any of the above can be a CPU, a microprocessor, an ASIC, or one or more integrated circuits for controlling the execution of the programs of the above feedback information transmission method.

[0176] In a possible design, the chip system further includes a memory, which is configured to store necessary program instructions and data. The processor and the memory can be decoupled and arranged on different devices, and connected through wired or wireless means to support the chip system to implement various functions in the above embodiments. Alternatively, the processor and the memory can be coupled on the same device.

[0177] Optionally, the computer instructions are stored in the memory.

[0178] Optionally, the memory is a storage unit in the chip, such as a register, a cache, etc. The memory can also be a storage unit in the terminal located outside the chip, such as a ROM or other type of static storage device that can store static information and instructions, a RAM, etc.

[0179] It can be understood that the memory in the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories.

[0180] The non-volatile memory can be a ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically EPROM (EEPROM), or a flash memory.

[0181] The volatile memory can be a RAM used as an external cache. RAM has many different types, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synch link DRAM (SLDRAM), and direct Rambus RAM.

[0182] The embodiments of the present disclosure are all the preferred embodiments of the present application, and are not intended to limit the protection scope of the present application, so that: any equivalent changes made according to the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. An AI-based content generation method, characterized by, The method comprises: obtaining a separated image and a background image and separating a feature object from the separated image; establishing a first light model belonging to the separated image and a second light model belonging to the background image; restoring the feature object using the first light model to obtain a basic feature object; processing the basic feature object using the second light model to obtain a fused feature object; transferring the fused feature object to the background image and dividing a fusion domain at a surrounding edge of the fused feature object; processing the fusion domain using a generative adversarial network to fuse the fused feature object with the background image; restoring the feature object using the first light model to obtain a basic feature object comprises: adjusting the first light model to make light sources in the first light model consistent and establishing a restoration direction line based on the adjusted first light model; calculating an enhancement or reduction degree on the restoration direction line; adjusting the feature object using the enhancement or reduction degree to obtain the basic feature object; processing the basic feature object using the second light model to obtain a fused feature object comprises: determining a brightness parameter of a transfer position of the fused feature object on the background image; adjusting the basic feature object using the brightness parameter; adjusting the basic feature object again using the second light model; obtaining an adjusted feature object, which is recorded as the fused feature object. 2.The AI-based content generation method of claim 1, wherein, establishing a first light model belonging to the separated image comprises: selecting a plurality of regions on the feature object after gray processing, each region comprising at least one pixel point; establishing a reference line using the regions; rotating the reference line and calculating pixel value change trends on the reference line to make at least two reference lines with the same pixel value change trend parallel or intersecting; establishing the first light model using the rotated reference line. 3.The AI-based content generation method of claim 1, wherein, After calculating the enhancement or reduction degree on the restoration direction line, further comprising: determining a cross-influence area and a blank-influence area of the restoration direction line; determining a restoration direction line associated with the blank-influence area; creating a secondary restoration direction line using the associated restoration direction line and adjusting the blank-influence area based on the secondary restoration direction line. 4.The AI-based content generation method of claim 1, wherein, Determining a brightness parameter of a transfer position of the fused feature object on the background image comprises: sequentially establishing a plurality of annular analysis reference lines inside the feature object and transferring the plurality of annular analysis reference lines to the background image, with a previous annular analysis reference line inside a next annular analysis reference line in a direction close to an edge of the feature object; calculating pixel value changes on the annular analysis reference lines and determining a brightest point and a dimmest point of light according to the pixel value changes; creating the brightness parameter using the brightest point and the dimmest point of light. 5.The AI-based content generation method of claim 1, wherein, When processing the fusion domain using the generative adversarial network, further comprising adjusting a width of the fusion domain. 6.An AI-based content generation apparatus, characterized by comprising: The method comprises: a first processing unit for obtaining a separated image and a background image and separating a feature object from the separated image; a light model establishing unit for establishing a first light model belonging to the separated image and a second light model belonging to the background image; a second processing unit for restoring the feature object using the first light model to obtain a basic feature object; a third processing unit for processing the basic feature object using the second light model to obtain a fused feature object; The fourth processing unit is configured to transfer the fused feature object to the background image and divide the fusion domain at a surrounding edge of the fused feature object; The fusion processing unit is configured to process the fusion domain using a generative adversarial network to fuse the fused feature object with the background image; The first light model is adjusted so that light sources in the first light model are consistent, and a restoration direction line is established based on the adjusted first light model; The intensity or reduction degree on the restoration direction line is calculated; The feature object is adjusted using the intensity or reduction degree to obtain the basic feature object; The second light model is used to process the basic feature object to obtain the fused feature object, including: Determining a luminance parameter of a transfer position of the fused feature object on the background image; Adjusting the basic feature object using the luminance parameter; Adjusting the basic feature object again using the second light model; An adjusted feature object is obtained, and the adjusted feature object is recorded as the fused feature object. The system includes:

7. A content generation system characterized by comprising: One or more memories for storing instructions; and One or more processors for calling and running the instructions from the memories to perform the method of any one of claims 1 to 5. The computer readable storage medium includes:

8. A computer-readable storage medium, characterized in that, A program, when the program is run by a processor, the method of any one of claims 1 to 5 is performed. ​

Images