AIGC-based Creative Material Correction System

Abstract

The present invention relates to the technical field of image generation, and particularly to a creative material correction system based on AIGC, including: a model disassembler that determines corresponding active model cores and a number of follower model cores according to the structure of the original image; a path constructor that determines corresponding test paths and stable jitter intervals according to the running paths of the active models; a defect diagnoser that is used to collect radial jitter corresponding to radial vibration and torsional jitter corresponding to torsional vibration; a material corrector that is used to repair the corresponding faulty frames; The present invention determines the contact situation between each follower model and the active model according to the vibration path and vibration amplitude, thereby determining a frame with low picture quality, effectively reducing the monitoring complexity while reducing the interference to the remaining images, thus effectively improving the positioning accuracy of the material correction system.

Description

Technical Field

[0001] The present invention relates to the technical field of image generation, and in particular, to a creative material correction system based on AIGC. Background Art

[0002] AIGC (Artificial Intelligence Generated Content) is widely used in the creation fields such as literature, graphics, music, and drama. It can generate high-quality works and is currently in a stage of rapid development and application. Through deep learning models, AIGC technology can automatically create content such as text, images, and videos, greatly improving the content production efficiency and creating content with high-quality creativity.

[0003] Its application in the advertising field is also becoming increasingly widespread. It not only improves the efficiency of advertising creativity but also optimizes advertising effects through data analysis and machine learning.

[0004] However, during the generation process, inevitable image conflicts occur between the colliding elements, and such problems have also become one of the main problems to be solved when generating content with AIGC.

[0005] The patent with the Chinese patent authorization publication number CN117115843B provides a convention and exhibition project content production system based on AIGC technology, including: a creative input unit, which is a platform for inputting creative design information; a creative production unit, which identifies the creative design information and processes it to generate initial material pictures; an effect evaluation unit, which corrects the initial material pictures to generate corrected materials; a creative adjustment unit, which makes detailed adjustments to the corrected materials; a server, which is connected to the creative input unit, the creative production unit, and the creative adjustment unit, and the server conducts data information transmission and final material output for the connected units. The present invention also provides a method for producing convention and exhibition project content, which is convenient for designers to master this system. This invention can learn the content of the material library through AI to create matching creative materials, improve the production efficiency of convention and exhibition design materials, and greatly shorten the working cycle of convention and exhibition projects.

[0006] The patent with the Chinese patent authorization announcement number CN110298229B discloses a video image processing method and device. Among them, the method includes: auditing the first video file and the second video file based on video clarity to obtain an audit result. The first video file is used for editing and processing, and the second video file is used for material sampling; performing face detection on the first video file to obtain a first face image, and performing face detection on the second video file to obtain a second face image; a feature extraction step, performing face feature extraction on the first face image to obtain first face image features, and performing face feature extraction on the second face image to obtain second face image features; performing feature mapping on the first face features and the second face features; replacing the first face features with the second face features to obtain a feature-replaced first face image. The invention improves the image clarity of the synthesized video.

[0007] For dynamic creative materials, there will be collisions between dynamic elements or mutual influences between paths in the materials. During the processing of generative AI, it is difficult to perform correct material generation or correction. Summary of the Invention

[0008] To this end, the present invention provides a creative material correction system based on AIGC to overcome the problem that in the prior art, there will be collisions between certain elements or mutual influences between paths in video materials, and it is difficult to perform correct material generation or correction during the processing of generative AI, resulting in a decline in the quality of generative materials.

[0009] To achieve the above object, the present invention provides a creative material correction system based on AIGC, which collects the original image materials of AIGC and corrects the materials according to the input corresponding text instructions, including:

[0010] A model disassembler, which determines the corresponding active model core and several follower model cores according to the structure of the original image, and determines several running paths of the active model and / or the follower model according to the text instructions;

[0011] A path builder, which determines the corresponding test path and the stable jitter interval according to the running path of the active model, and,

[0012] According to the coordinate systems of each follower model, determines the directions corresponding to the radial vibration and torsional vibration when the follower model collides with the active model in the test path;

[0013] A defect diagnoser, which is used to collect the radial jitter corresponding to the radial vibration and the torsional jitter corresponding to the torsional vibration, and determines the mutation phase and the mutation path according to the test path, and,

[0014] Responds to the mutation phase to determine the potential fault frame of the active model or the follower model, or,

[0015] Determine the potential fault displacement of the active model in response to the mutation path;

[0016] A material corrector that determines an actual fault frame according to the potential fault frame and / or the potential fault displacement, determines a corresponding stable operation path according to the previous frame and / or the next frame of the actual fault frame, and repairs the corresponding fault frame;

[0017] Wherein, the stable jitter interval is related to the accuracy of the original image, and is mapped to the jitter bandwidth of the radial jitter or the torsional jitter.

[0018] Furthermore, the path builder determines the number of collision events of the original image according to the number of frames of the active model and the respective slave models, and determines the collision paths of the respective slave models according to the number of collision events;

[0019] Record the maximum amplitude of the radial vibration or torsional vibration of each slave model as the radial jitter or torsional jitter of a single collision event according to the collision path;

[0020] Wherein, the amplitude of the radial jitter is the maximum vibration distance between the core of the slave model and the core of the active model;

[0021] The amplitude of the torsional jitter is the maximum vibration distance at which the core of the slave model deviates during operation.

[0022] Furthermore, the model disassembler sets an active model core corresponding to the active model according to the geometric position of the active model, and determines a corresponding slave model core according to the corresponding position and angle of the slave model;

[0023] In a single collision, the model disassembler sets a slave model coordinate system for each slave model based on the contact position between the active model and the respective slave models with the core of the slave model as a reference.

[0024] Furthermore, the defect diagnoser is provided with a stable jitter interval, and sets a plurality of jitters exceeding the stable jitter interval as mutation jitters;

[0025] Determine the projection of the mutation jitter on the corresponding plane where the line connecting the core of the slave model and the core of the active model is located as the mutation phase of the mutation jitter according to the position of each mutation jitter on the slave model.

[0026] Furthermore, for a single mutation path, the defect diagnoser also determines the equivalent mutation of each slave model according to the mutation phase, determines the corresponding path for the recurrence of each equivalent mutation, and records it as the mutation path of the mutation phase.

[0027] Further, the defect diagnoser compares the mutation path with the test path to form a corresponding mutation fitting, determines the active model boundary of the active model and the follower model boundary of the follower model according to the mutation fitting, and records the active model boundary and the follower model boundary as potential fault displacements.

[0028] Further, the defect diagnoser maps the mutation path onto the test path to form a corresponding mutation mapping relationship, and records several frames in which the follower model contacts the active model in the collision path where the mutation jitter occurs as potential fault frames according to the mutation mapping relationship.

[0029] Further, when the potential fault frames and the potential fault displacements overlap in several test paths, the material trimmer determines the potential fault frames of the corresponding active model as actual fault frames, or,

[0030] when the potential fault frames and the potential fault displacements do not overlap in any test path, it is determined that the follower model has a fault;

[0031] Wherein, a single test includes at least 2 consecutive test paths.

[0032] Further, if the mutation path does not overlap with the test path within any test path, the model disassembler determines that the collision path is abnormal and adjusts the structure of the original image.

[0033] Further, when determining the actual fault frames, the defect diagnoser performs fault frame verification and tests at least 3 test paths;

[0034] Wherein, each test path marks the actual fault frame, and when running to each follower model, verification is performed according to the interaction between each follower model and the active model.

[0035] Compared with the prior art, the beneficial effects of the present invention are as follows: by detecting the vibrations of the original image and each follower model, and determining the contact situation between each follower model and the active model according to the vibration path and vibration amplitude, a frame with low picture quality is determined, effectively reducing the monitoring complexity while reducing the interference to the remaining images, thereby effectively improving the positioning accuracy of the material correction system.

[0036] Further, by determining the phenomenon of image distortion that may be caused by model jitter, while effectively reducing the number of image recognitions for material correction, the targeting of the corresponding positions of distorted images is improved, thereby further enhancing the positioning accuracy of the material correction system.

[0037] Furthermore, by setting a reliable stable jitter interval, it avoids repairing unobservable images and thus affecting the overall progress of the repair operation. At the same time, it avoids the overfitting problem caused by the single factor material of the correction system, effectively improving the efficiency of material correction while enhancing the display quality of material correction.

[0038] Furthermore, by reproducing the path of the material movement to verify the material repair effect and optimizing the material according to the verification result, it effectively improves the accuracy of the video material repair by the material repair system while enhancing the display quality of the material. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Connection schematic diagram of the creative material correction system based on AIGC of the present invention;

[0040] Figure 2 Operation flow chart of the creative material correction system based on AIGC of the present invention;

[0041] Figure 3 Schematic diagram of model definition of the active model and the follower model of the embodiment of the present invention;

[0042] Figure 4 Collision schematic diagram of the active model and the follower model of the embodiment of the present invention;

[0043] Wherein: 1, active model; 11, active model core; 12, active model moving direction; 13, active model radial direction; 14, active model torsion direction; 2, follower model; 21, follower model core; 22, follower model moving direction; 23, follower model radial direction; 24, follower model torsion direction; 3, model boundary; 31, abnormal model boundary. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] In order to make the objectives and advantages of the present invention clearer, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0045] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and do not limit the protection scope of the present invention.

[0046] It should be noted that in the description of the present invention, the terms indicating directions or positional relationships such as "upper", "lower", "left", "right", "inner", "outer", etc. are based on the directions or positional relationships shown in the drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present invention.

[0047] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installation", "connection", and "coupling" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0048] Please refer to Figure 1 and Figure 2 as shown, which are the connection schematic diagram of the AIGC-based creative material correction system of the present invention and the operation flowchart of the AIGC-based creative material correction system of the present invention. Before correcting the AIGC creative material, it collects the original image material of AIGC and corrects the material according to the input corresponding text instructions, including:

[0049] The model disassembler determines the corresponding active model core and several follower model cores according to the structure of the original image, and determines several running paths of the active model and / or the follower models according to the text instructions;

[0050] The path builder determines the corresponding test path and the stable jitter interval according to the running path of the active model, and,

[0051] According to the coordinate systems of the respective follower models, determines the directions corresponding to the radial vibration and torsional vibration when the follower model collides with the active model in the test path;

[0052] The defect diagnoser is used to collect the radial jitter corresponding to the radial vibration and the torsional jitter corresponding to the torsional vibration, and determine the mutation phase and the mutation path according to the test path, and,

[0053] Responds to the mutation phase to determine the potential fault frame of the active model or the follower model, or,

[0054] Responds to the mutation path to determine the potential fault displacement of the active model;

[0055] A material corrector that determines the actual fault frame based on potential fault frames and / or potential fault displacements, determines the corresponding smooth operation path based on the previous frame and / or the next frame of the actual fault frame, and repairs the corresponding fault frame;

[0056] Among them, the smooth jitter interval is related to the precision of the original image, and it is mapped to the jitter bandwidth of radial jitter or torsional jitter.

[0057] In implementation, jitter is the image blurring phenomenon caused by vibration in the corresponding direction or rotational direction, specifically manifested as "ghosting" composed of several overlapping images at a certain position.

[0058] By detecting the vibrations of the original image and each follower model, determining the contact situation between each follower model and the active model according to the vibration path and vibration amplitude, thereby determining a frame with low picture quality, effectively reducing the monitoring complexity while reducing the interference to the remaining images, thus effectively improving the positioning accuracy of the material correction system.

[0059] For example, in AR advertisement production, the original image material may include a product model and a real-world background. The model disassembler will identify the product model as the core of the active model and the background elements as the core of the follower models.

[0060] The path builder will determine the test path and the smooth jitter interval according to the movement path of the product model in the advertisement, and calculate the collisions and vibrations between the product model and the background elements.

[0061] The defect diagnostician will monitor the jitter in the fusion of the product model and the background to ensure the naturalness and attractiveness of the advertisement. If abnormal jitter is detected, it will determine the mutation phase and mutation path, and identify potential fault frames or displacements.

[0062] The material corrector will determine the actual fault location according to the identified fault frames and displacements, and repair or optimize the fusion of the product model and the background to ensure the visual effect of the advertisement.

[0063] Specifically, the path builder determines the number of collision events of the original image according to the number of frames of the active model and each follower model, and determines the collision path of each follower model according to the number of collision events;

[0064] Record the maximum amplitude of the radial vibration or torsional vibration of each follower model as the radial jitter or torsional jitter of a single collision event according to the collision path;

[0065] Among them, the amplitude of the radial jitter is the maximum vibration distance between the core of the follower model and the core of the active model;

[0066] The amplitude of the torsional jitter is the maximum vibration distance at which the core of the follower model deviates during operation.

[0067] By utilizing the phenomenon of image distortion that may be caused by determining model jitter, while effectively reducing the number of image recognitions for material correction, the targeting of the corresponding positions of distorted images is improved, thereby further enhancing the positioning accuracy of the material correction system.

[0068] Example 1

[0069] In an advertising plot, there is a scene where the protagonist (active model) runs in the forest, avoiding falling rocks (follow - up model). This scene consists of 300 frames, with each frame representing an instantaneous action and 30 frames per second.

[0070] The path builder analyzes the running path of the protagonist and the falling path of the rocks, and determines that there are 20 potential collision events in 300 frames.

[0071] For each collision event, the path builder records the maximum radial maximum vibration distance between the falling rocks (follow - up model) and the protagonist (active model). For example, in the 150th frame, the maximum amplitude of the distance between a falling rock and the protagonist is 5 pixels, which is considered the radial jitter amplitude of this collision event.

[0072] Example 2

[0073] In the simulation of a robot soccer game, a robot (active model) attempts to dribble through the opposing defensive robot (follow - up model). The entire breakthrough process is simulated for 120 frames, with 24 frames per second.

[0074] The path builder analyzes the path of the dribbling robot and the movement of the defensive robot, and determines that there are 15 potential collision events in 120 frames.

[0075] For each collision event, the path builder records the maximum radial maximum vibration distance between the defensive robot and the dribbling robot. For example, in the 75th frame, the maximum radial maximum vibration distance of a collision is 3 centimeters, which is considered the radial jitter amplitude of this collision event.

[0076] Example 3

[0077] In an aerospace simulation, a satellite (active model) moves in space and needs to avoid colliding with other space debris (follow - up model). The simulation process consists of 500 frames, with 16 frames per second.

[0078] The path builder analyzes the orbit of the satellite and the trajectory of the space debris, and determines that there are 30 potential collision events in 500 frames.

[0079] For each collision event, the path builder records the maximum radial maximum vibration distance between space debris and the satellite. For example, at the 250th frame, the maximum amplitude of the distance between a piece of space debris and the satellite is 0.5 meters, which is considered the radial jitter amplitude of this collision event.

[0080] Specifically, the model disassembler sets the active model core corresponding to the active model according to the geometric position of the active model, and determines the corresponding follower model core according to the position and angle corresponding to the follower model;

[0081] In a single collision, the model disassembler sets the follower model coordinate system of each follower model based on the follower model core according to the contact position between the active model and each follower model.

[0082] Compared with the above-mentioned Embodiment 1, the corresponding active model core is the centroid of the protagonist, and the follower model core is the centroid of each falling rock;

[0083] For the above-mentioned Embodiment 2, the corresponding active model core is the geometric center of the ball, and the follower model core is the centroid of each robot;

[0084] For the above-mentioned Embodiment 3, the corresponding active model core is the center of the satellite, and the follower model core is the geometric center of each piece of space debris;

[0085] It can be understood that the core of the active model and the core of the follower model should be correspondingly changed according to the main body in the AIGC-generated image, and the one with a higher frequency of appearance in the entire material should be selected as the active model, and the remaining models as the follower models.

[0086] Please refer to Figure 3 As shown, it is a schematic diagram of the model definition of the active model and the follower model of the embodiment of the present invention. In the figure, both the active model 1 and the follower model 2 are ellipsoids, and their centers of gravity are the active model core 11 and the follower model core 21 respectively;

[0087] Among them, the mapping of the active model moving direction 12 of the active model 1 on the active model core 11 is the active model radial 13, and its rotation direction around the active model 1 is the active model torsion direction 14;

[0088] The mapping of the follower model moving direction 22 of the follower model 2 on the follower model core 21 is the follower model radial 23, and its rotation direction around the follower model 2 is the follower model torsion direction 24;

[0089] It is not difficult to understand that by specifying the above parameters, the moving modes of the active model 1 and the follower model 2 in a continuous process can be described, and the effects generated during their collision can also be better reflected.

[0090] Specifically, the defect diagnoser is provided with a stable jitter range, and several jitters exceeding the stable jitter range are set as mutation jitters;

[0091] According to the positions of the mutation jitters on the follow-up model, the projections of the mutation jitters on the corresponding plane where the line connecting the core of the follow-up model and the core of the active model is located are determined as the mutation phases of the mutation jitters.

[0092] The defect diagnoser determines the mutation phase of the advertisement material according to the performance of the advertisement material on different platforms. It is not difficult to understand that the generation of this mutation phase stems from the image processing capabilities and display performances of different platforms:

[0093] For example, on the mobile phone platform, the mobile phone screens are mostly 3.5 - 6.8 inches. At this time, when the advertisement is displayed, users can observe fewer image details, but more elements can be stacked;

[0094] For example, on the PC (personal computer) platform, the conventional monitor size is 19 - 35 inches. At this time, when the advertisement is displayed, users can observe more image details, and the number of elements that can be stacked is moderate;

[0095] It is not difficult to see that the number of mutation phases of the same material on the mobile phone platform is smaller, and the number of mutation phases on the PC platform is larger;

[0096] When implementing, the stable jitter range related to the mutation phase can be set according to the resolution. For example:

[0097] For an advertisement with a size of 1m × 0.6m, which contains a material with a size of 0.25m × 0.25m, for a 720P resolution, the corresponding stable jitter range can be 10cm;

[0098] For a 1080P resolution, the corresponding stable jitter range can be 8cm;

[0099] When the display is clearer, the abnormal movement of the material will be amplified. Therefore, the higher the resolution of the advertisement material, the smaller the corresponding stable jitter range.

[0100] Specifically, for a single mutation path, the defect diagnoser also determines the equivalent mutations of each follow-up model according to the mutation phase, determines the corresponding paths reproduced by each equivalent mutation, and records them as the mutation paths of this mutation phase.

[0101] By setting a reliable stable jitter range, it avoids repairing unobservable images and thus affecting the overall progress of the repair operation. At the same time, it avoids the overfitting problem caused by the single factor material of the correction system. While effectively improving the correction efficiency of the material, it improves the display quality of the material correction.

[0102] Specifically, the defect diagnoser compares the mutation path with the test path to form a corresponding mutation fit, determines the active model boundary of the active model and the follower model boundary of the follower model according to the mutation fit, and records the active model boundary and the follower model boundary as potential fault displacements.

[0103] Specifically, the defect diagnoser maps the mutation path onto the test path to form a corresponding mutation mapping relationship, and records several frames in which the follower model contacts the active model in the collision path with mutation jitter as potential fault frames according to the mutation mapping relationship.

[0104] In implementation, using a large AI model can optimize the placement effect of advertising materials:

[0105] In the dynamic analysis of advertising materials, the defect diagnoser sets a stable jitter interval to identify the normal and abnormal performances of the materials on different placement platforms. Performances exceeding this interval are marked as mutation jitters.

[0106] Specifically, when the potential fault frames and the potential fault displacements overlap in several test paths, the material trimmer determines the potential fault frames of the corresponding active model as actual fault frames, or,

[0107] When the potential fault frames and the potential fault displacements do not overlap in any test path, it is determined that the follower model has a fault;

[0108] Among them, a single test includes at least 2 consecutive test paths.

[0109] In implementation, when a fault occurs in the active model or the follower model, it is determined that the boundary definition of the active model or the follower model is abnormal, and the corresponding boundary is regenerated. Please refer to Figure 4 As shown, it is a schematic diagram of the collision between the active model and the follower model in an embodiment of the present invention. In the figure, the active model 1 and the follower model 2 collide in a certain frame. When generating this collision, an abnormal model boundary 31 is generated. At this time, the abnormal model boundary 31 is corrected and smoothed to form the model boundary 3.

[0110] Specifically, if the mutation path does not overlap with the test path within any test path, the model disassembler determines that the collision path is abnormal and adjusts the structure of the original image.

[0111] Specifically, when determining the actual fault frames, the defect diagnoser performs fault frame verification and tests at least 3 test paths;

[0112] Among them, each test path marks the actual fault frames, and when running to each follower model, verification is performed according to the interaction between each follower model and the active model.

[0113] Verify the repair effect of the material by replicating the path of the material movement, and optimize the material according to the verification result, which not only effectively improves the accuracy of the material repair system for video material repair, but also improves the display quality of the material.

[0114] So far, the technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0115] The above are only the preferred embodiments of the present invention and are not used to limit the present invention; for those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A creative material correction system based on AIGC, which collects original image materials of AIGC and corrects the materials according to input corresponding text instructions, characterized in that: include: A model disassembler determines the corresponding active model core and several follower model cores according to the structure of the original image, and determines several running paths of the active model and / or the follower model according to the text instructions; A path builder determines the corresponding test path and stable jitter interval according to the running path of the active model, and According to the coordinate system of each follower model, the directions of radial vibration and torsional vibration corresponding to the collision between the follower model and the active model in the test path are determined; A defect diagnostic device, which is used to collect radial jitter corresponding to radial vibration and torsional jitter corresponding to torsional vibration, and determine the mutation phase and mutation path according to the test path, and, determining a potential fault frame of an active model or a follower model in response to the sudden phase change, or, determining a potential fault displacement of an active model in response to the mutation path; A material corrector, which determines an actual fault frame according to the potential fault frame and / or the potential fault displacement, determines a corresponding smooth running path according to a previous frame and / or a subsequent frame of the actual fault frame, and repairs the corresponding fault frame; The stable jitter interval is related to the accuracy of the original image, and is mapped to the jitter bandwidth of the radial jitter or the torsional jitter; The path builder determines the number of collision events of the original image according to the number of frames of the active model and each follower model, and determines the collision path of each follower model according to the number of collision events; The maximum amplitude of the radial vibration or torsional vibration of each follow-up model is recorded according to the collision path as the radial jitter or torsional jitter of a single collision event; Wherein, the amplitude of the radial jitter is the vibration distance between the follower model core and the active model core; The amplitude of the torsional vibration is the vibration distance that the follower model core deviates during operation.

2. The creative material correction system based on AIGC according to claim 1, characterized in that: The model disassembler sets an active model core corresponding to the active model according to the geometric position of the active model, and determines a corresponding follower model core according to the position and angle corresponding to the follower model; In a single collision, the model disassembler sets the follower model coordinate system of each follower model based on the follower model core according to the contact position between the active model and each follower model.

3. The creative material correction system based on AIGC according to any one of claims 1 or 2, characterized in that: The defect diagnostic device is provided with a stable jitter interval, and sets a number of jitters exceeding the stable jitter interval as sudden jitters; According to the position of each sudden jitter on the follower model, the projection of the sudden jitter on the corresponding surface where the line connecting the follower model core and the active model core is located is determined as the sudden phase of the sudden jitter.

4. The AIGC-based creative material correction system according to claim 3, characterized in that: For a single mutation path, the defect diagnostic device further determines the equivalent mutations of each follow-up model according to the mutation phase, and determines the corresponding path for the recurrence of each equivalent mutation, and records it as the mutation path of the mutation phase.

5. The AIGC-based creative material correction system according to claim 4, characterized in that: The defect diagnostic device compares the mutation path with the test path to form a corresponding mutation fit, determines an active model boundary in the active model and a follower model boundary in the follower model according to the mutation fit, and records the active model boundary and the follower model boundary as potential fault displacements.

6. The creative material correction system based on AIGC according to claim 4, characterized in that: The defect diagnostic device maps the mutation path onto the test path and forms a corresponding mutation mapping relationship, and based on the mutation mapping relationship, records several frames in which the follower model contacts the active model in the collision path where the mutation jitter occurs as potential fault frames.

7. The creative material correction system based on AIGC according to claim 5 or 6, characterized in that: The material modifier determines the potential fault frame of the corresponding active model as an actual fault frame in response to the potential fault frame and the potential fault displacement overlapping in a plurality of test paths, or, In response to the potential fault frame and the potential fault displacement not overlapping in any test path, determining that the follower model fails; A single test includes at least two consecutive test paths.

8. The AIGC-based creative material correction system according to claim 7, characterized in that: If the mutation path does not overlap with the test path in any test path, the model disassembler determines that the collision path is abnormal and adjusts the structure of the original image.

9. The creative material correction system based on AIGC according to claim 8, characterized in that: When determining the actual fault frame, the defect diagnostic device performs fault frame verification and performs tests on at least three test paths; Each test path marks the actual fault frame, and when running to each follower model, performs verification according to the interaction between each follower model and the active model.

Images