Visual model training data set expansion method and device, equipment and storage medium
By obtaining the shooting information in the video stream and 3D models to generate two-dimensional images, the motion trajectory data set of moving objects is expanded, the problem of insufficient sample data is solved, and the model training effect is improved.
Patent Information
- Application Number
- CN202411091429.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-09
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-08-09
AI Technical Summary
In the process of predictive model training that requires prediction of the motion path of moving objects, it is difficult to obtain enough sample data, resulting in poor model training results.
By obtaining the shooting information of the specified moving objects in the video stream, a specified motion path is generated, and a two-dimensional image of each point is obtained based on the 3D model, and added to the video stream to expand the training data set.
The training data set is effectively expanded, the quantity and quality of sample data are increased, and the training effect of the model is improved.
Smart Images

Figure CN119048859A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of image processing, and in particular to a method, device and storage medium for expanding a visual model training data set. Background Art
[0002] In some application scenarios, for example, in the training process of a prediction model that needs to predict the motion path of a moving object, a large amount of sample data is needed to train the model. However, due to some reasons, not enough sample data can be obtained. Therefore, in order to ensure that the model can be effectively trained, the existing sample data needs to be expanded. Summary of the invention
[0003] In view of the above technical problems, the technical solution adopted by the present invention is:
[0004] According to a first aspect of the present invention, a method for expanding a visual model training data set is provided, the method comprising the following steps:
[0005] S100, obtaining a video stream to which a specified motion object is currently to be added as a target video stream.
[0006] S200, generating a designated motion path in a scene corresponding to the target video stream, and setting attribute information corresponding to the designated motion path, wherein the attribute information includes motion time and motion form.
[0007] S300, obtaining shooting information of each point on the specified motion path, wherein the shooting information includes a shooting angle and a shooting distance, wherein the shooting angle of a certain point refers to the angle between the line between the point and the corresponding shooting device and the shooting sight line of the shooting device, and the shooting distance of a certain point refers to the distance between the point and the corresponding shooting device; obtaining a shooting information set C corresponding to the motion path = {C1, C2, ..., C i , ..., C n}, the i-th shooting information C in C i is the shooting information of the i-th point on the specified motion path, C i =(α i , d i ), i ranges from 1 to n, n is the number of points on the specified motion path, α i is the shooting angle corresponding to the i-th point, d i is the shooting distance corresponding to the i-th point.
[0008] S400, based on the image set to which the specified moving object needs to be added in the target video stream, based on C and the 3D model corresponding to the specified moving object, obtain the two-dimensional image corresponding to each point of the specified moving object, and put the obtained two-dimensional image into the corresponding position of the corresponding image in the target video stream.
[0009] According to a second aspect of the present invention, a device for expanding a visual model training data set is provided, the device comprising:
[0010] The first acquisition module is used to acquire the video stream to which the specified motion object is currently added as the target video stream.
[0011] The path generation module is used to generate a specified motion path in the scene corresponding to the target video stream, and set attribute information corresponding to the specified motion path, wherein the attribute information includes motion time and motion form.
[0012] The second acquisition module is used to acquire the shooting information of each point on the specified motion path, wherein the shooting information includes a shooting angle and a shooting distance, wherein the shooting angle of a certain point refers to the angle between the line between the point and the corresponding shooting device and the shooting sight line of the shooting device, and the shooting distance of a certain point refers to the distance between the point and the corresponding shooting device; obtain the shooting information set C corresponding to the motion path = {C1, C2, ..., C i , ..., C n}, the i-th shooting information C in C i is the shooting information of the i-th point on the specified motion path, C i =(α i , d i ), i ranges from 1 to n, n is the number of points on the specified motion path, α i is the shooting angle corresponding to the i-th point, d i is the shooting distance corresponding to the i-th point.
[0013] An adding module is used to obtain the two-dimensional image corresponding to each point of the specified moving object based on the image set to which the specified moving object needs to be added in the target video stream, based on C and the 3D model corresponding to the specified moving object, and put the obtained two-dimensional image into the corresponding position of the corresponding image in the target video stream.
[0014] According to a third aspect of the present invention, there is provided an electronic device comprising a processor and a memory; the processor is used to execute the steps of the method described in the first aspect of the present invention by calling a program or instruction stored in the memory.
[0015] According to a fourth aspect of the present invention, there is provided a non-transitory computer-readable storage medium storing a program or instructions, wherein the program or instructions enable a computer to execute the steps of the method described in the first aspect of the present invention.
[0016] The present invention has at least the following beneficial effects:
[0017] The method for expanding a visual model training data set provided by an embodiment of the present invention can add motion trajectories of moving objects to a training data set, and can expand sample data.
[0018] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0020] Figure 1 A flowchart of a method for expanding a visual model training data set provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.
[0023] It should be noted that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the steps as sequential processes, many of the steps can be implemented in parallel, concurrently or simultaneously. In addition, the order of the steps can be rearranged. The process can be terminated when its operation is completed, but it can also have additional steps not included in the accompanying drawings. The process can correspond to a method, function, procedure, subroutine, subprogram, etc.
[0024] An embodiment of the present invention provides a method for expanding a visual model training data set, the method comprising: Figure 1 The following steps are shown:
[0025] S100, obtaining a video stream to which a specified motion object is currently to be added as a target video stream.
[0026] In the embodiment of the present invention, the target video stream may be a video stream specified by the user, or may be a video stream obtained by continuously shooting a corresponding scene, for example, a video stream obtained by continuously shooting a traffic intersection, or a video stream obtained by continuously shooting a certain section of road.
[0027] In the embodiment of the present invention, the designated moving object may be a moving object, such as a pedestrian, a vehicle, an airplane, etc.
[0028] S200, generating a designated motion path in a scene corresponding to the target video stream, and setting attribute information corresponding to the designated motion path, wherein the attribute information includes motion time and motion form.
[0029] In the embodiment of the present invention, the specified motion path may be any path specified by the user. In specific implementation, it may be specified in the first frame image of the target video stream.
[0030] In the embodiment of the present invention, the motion time refers to the time required from the starting position to the end position of the specified motion path, which can be the time specified by the user. The motion form is the motion speed type from the starting position to the end position of the specified motion path, including uniform motion, accelerated motion and other forms. In an exemplary embodiment of the present invention, the motion form can be uniform motion.
[0031] S300, obtaining shooting information of each point on the specified motion path, wherein the shooting information includes a shooting angle and a shooting distance, wherein the shooting angle of a certain point refers to the angle between the line between the point and the corresponding shooting device and the shooting sight line of the shooting device, and the shooting distance of a certain point refers to the relative distance between the point and the corresponding shooting device; obtaining a shooting information set C corresponding to the motion path = {C1, C2, ..., C i , ..., C n}, the i-th shooting information C in Ci is the shooting information of the i-th point on the specified motion path, C i =(α i , d i ), i ranges from 1 to n, n is the number of points on the specified motion path, α i is the shooting angle corresponding to the i-th point, d i is the shooting distance corresponding to the i-th point.
[0032] Since the position and sight line of the camera are relatively fixed during shooting, the pixel points in each frame of the video stream are basically unchanged relative to the shooting angle and shooting distance of the camera. Therefore, the shooting angle and shooting distance of each pixel point in the image can be obtained based on existing knowledge.
[0033] S400, based on the image set to which the specified moving object needs to be added in the target video stream, based on C and the 3D model corresponding to the specified moving object, obtain the two-dimensional image corresponding to each point of the specified moving object, and put the obtained two-dimensional image into the corresponding position of the corresponding image in the target video stream.
[0034] Further, in an exemplary embodiment, S400 may specifically include:
[0035] S410, set variable counter j=1.
[0036] S420, if j≤m, execute S430, otherwise, execute S470; m is the number of frames of the image in the image set;
[0037] S430: for the j-th frame image IMG in the image set, j , get IMG j and the previous frame image IMG j-1 The time interval △t j(j-1) .
[0038] In the embodiment of the present invention, each frame of image corresponds to a shooting time, Δt j(j-1) for IMG j and IMG j-1 The difference between the shooting times.
[0039] S440, based on the starting position, end position, movement time, movement form, IMG corresponding to the specified movement path j-1 The corresponding added point position G j-1 and △t j(j-1) , get IMG j The position G of the adding point where the specified moving object needs to be added j ; G1 is the starting point position corresponding to the specified motion path.
[0040] It is known to those skilled in the art that, assuming that the specified moving object moves along a specified moving path, based on the starting position, end position, moving time, moving form, IMG corresponding to the specified moving path, j-1 The corresponding added point position G j-1 and △t j(j-1) , we can get the specified moving object from the previous position G j-1 Elapsed time △t j(j-1) After that, any method for obtaining the position of G1 in the specified motion path belongs to the protection scope of the present invention.
[0041] S450, based on G j , obtain the corresponding shooting information from C, and based on G j The corresponding shooting information controls the virtual camera corresponding to the 3D model to shoot the specified moving object in the 3D model, and obtains the corresponding two-dimensional image as G j The corresponding two-dimensional image DIMG j .
[0042] In the embodiment of the present invention, the virtual camera and the shooting device corresponding to the target video stream have the same parameters.
[0043] It is known to those skilled in the art that any j The corresponding shooting information controls the virtual camera corresponding to the 3D model to shoot the specified moving object in the 3D model, and the method of obtaining the corresponding two-dimensional image belongs to the protection scope of the present invention.
[0044] In the embodiment of the present invention, due to the G j The corresponding shooting information controls the virtual camera corresponding to the 3D model to shoot the specified moving object in the 3D model to obtain a corresponding two-dimensional image, thereby making the moving object added to the target video stream relatively real.
[0045] S460, from DIMG j The specified moving object is segmented out and the segmented specified moving object is added to IMG j G j Place.
[0046] In the embodiment of the present invention, the existing image superposition method can be used to add the segmented specified moving object to the IMG j G j Place.
[0047] S470, add the specified motion image to the IMG j Add to the current intermediate image set; set j=j+1, and execute S420; the initial value of the current intermediate image set is empty.
[0048] S480: Use the current intermediate image set as the target image set corresponding to the target video stream.
[0049] In another exemplary embodiment of the present invention, S400 may specifically include:
[0050] S411, based on C, generating a camera movement path corresponding to the virtual camera corresponding to the 3D model.
[0051] Those skilled in the art know that any method for generating a camera movement path corresponding to a virtual camera corresponding to the 3D model based on C falls within the protection scope of the present invention.
[0052] S412, based on the camera movement path, controlling the virtual camera to shoot, obtaining a corresponding shooting image set IMGS = {DIMG1, DIMG2, ..., DIMG i , ..., DIMG m}, DIMG i C based virtual camera i The two-dimensional image obtained.
[0053] S413, set variable counter j=1.
[0054] S414, if j≤m, execute S415, otherwise, execute S420; m is the number of frames of the image in the image set;
[0055] S415: for the j-th frame image IMG in the image set, j , get IMG j and the previous frame image IMG j-1 The time interval △t j(j-1) .
[0056] S416, based on the starting position, end position, movement time, movement form, IMG corresponding to the specified movement path j-1 The corresponding added point position G j-1 and △t j(j-1) , get IMG j The position G of the adding point where the specified moving object needs to be added j ; G1 is the starting point position corresponding to the specified motion path.
[0057] S417, Get from IMGS to IMG j The corresponding two-dimensional image, as IMG j The corresponding two-dimensional image DIMG j .
[0058] In the embodiment of the present invention, according to Gj Get it from IMGS j The corresponding two-dimensional image, as IMG j The corresponding two-dimensional image DIMG j .
[0059] S418, from DIMG j The specified moving object is segmented out and the segmented specified moving object is added to IMG j G j Place.
[0060] S419, add the specified motion image to the IMG j Add to the current intermediate image set; set j=j+1, and execute S414; the initial value of the current intermediate image set is empty.
[0061] S420: Use the current intermediate image set as the target image set corresponding to the target video stream.
[0062] The specific implementation of S413 to S416 may refer to the specific implementation of S410 to S440 described above, and the specific implementation of S418 to S420 may refer to the specific implementation of S460 to S480 described above.
[0063] Compared with the previous embodiments, this embodiment first generates a moving path of the shooting device based on C, and then shoots the 3D model according to the moving path, maps the 3D model into a two-dimensional space, and obtains a two-dimensional image set corresponding to each point on the specified motion path. In this way, when the corresponding two-dimensional image is needed, it can be directly obtained from the image set, which can improve efficiency.
[0064] Further, in the embodiment of the present invention, the center of the bounding box of the moving object is designated as the corresponding adding point.
[0065] Furthermore, in the embodiment of the present invention, after S460, the following steps are also included:
[0066] S462, for IMG j The target object in the image is identified to obtain a corresponding identification frame; the target object includes the specified moving object.
[0067] In an embodiment of the present invention, a target detection algorithm such as a YOLO series target detection algorithm may be used to identify a target object in an image. The specific type of the target object may be set based on actual needs.
[0068] S464, if DIMG j The identification box and IMG corresponding to the specified moving object segmented j There is an overlapping area between any recognition boxes r in dj <d r , indicating that the specified moving object is located in front of the target object corresponding to the identification frame relative to the shooting device, and the area belonging to the identification frame r in the overlapping area needs to be blocked. If d j >d r , indicating that the specified moving object is located behind the target object corresponding to the recognition frame relative to the shooting device, and the DIMG in the overlapping area needs to be j The area of the recognition box corresponding to the specified moving object segmented from the image is blocked, d r is the distance between the recognition frame r and the corresponding camera device, j ranges from 1 to f(i)-1, and f(i) is IMG j The number of identification boxes in ; execute S470.
[0069] In the embodiment of the present invention, any method for shielding a region in an image falls within the protection scope of the present invention.
[0070] Furthermore, in the embodiment of the present invention, after S418, the following steps are also included:
[0071] S4181, for IMG j The target object in the image is identified to obtain a corresponding identification frame; the target object includes the specified moving object.
[0072] S4182, if DIMG j The identification box and IMG corresponding to the specified moving object segmented j There is an overlapping area between any recognition boxes r in d j <d r , block the area in the overlapping area that belongs to the recognition box r. If d j >d r , the overlapping area belongs to DIMG j The area of the recognition box corresponding to the specified moving object segmented from the image is blocked, d r is the distance between the recognition frame r and the corresponding camera device, j ranges from 1 to f(i)-1, and f(i) is IMG j The number of identification boxes in ; execute S419.
[0073] The method for expanding the visual model training data set provided by an embodiment of the present invention, based on the relative distance and shooting angle of each point of the specified motion path of the scene corresponding to the target video stream to the shooting device, shoots the 3D model of the specified moving object, obtains the corresponding two-dimensional image, and then adds it to the corresponding image in the target video stream, thereby generating a video stream with the motion trajectory of the moving object added, and can expand the sample data used for model prediction.
[0074] Based on the same inventive concept, an embodiment of the present invention provides a device for expanding a visual model training data set, the device comprising:
[0075] The first acquisition module is used to acquire the video stream to which the specified motion object is currently added as the target video stream.
[0076] The path generation module is used to generate a specified motion path in the scene corresponding to the target video stream, and set attribute information corresponding to the specified motion path, wherein the attribute information includes motion time and motion form.
[0077] The second acquisition module is used to acquire the shooting information of each point on the specified motion path, wherein the shooting information includes the shooting angle and the shooting distance between the corresponding shooting device; and obtain the shooting information set C corresponding to the motion path = {C1, C2, ..., C i , ..., C n}, the i-th shooting information C in C i is the shooting information of the i-th point on the specified motion path, C i =(α i , d i ), i ranges from 1 to n, n is the number of points on the specified motion path, α i is the shooting angle corresponding to the i-th point, d i is the shooting distance corresponding to the i-th point.
[0078] An adding module is used to obtain the two-dimensional image corresponding to each point of the specified moving object based on the image set to which the specified moving object needs to be added in the target video stream, based on C and the 3D model corresponding to the specified moving object, and put the obtained two-dimensional image into the corresponding position of the corresponding image in the target video stream.
[0079] The device can be used to perform Figure 1 Therefore, for the functions that can be realized by each functional module of the device, please refer to Figure 1 The description of the illustrated embodiment is not repeated in detail.
[0080] An embodiment of the present invention also provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are configured to execute the method described in the embodiment of the present invention.
[0081] The embodiment of the present invention further provides a non-transitory computer-readable storage medium storing computer-executable instructions, wherein the computer instructions are used to execute the method described in the embodiment of the present invention.
[0082] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution disclosed in the present invention can be achieved, and this document does not limit this.
[0083] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for expanding a visual model training dataset. It is characterized in that The method comprises the following steps: S100, obtaining a video stream to which a specified moving object is currently to be added as a target video stream; S200, generating a designated motion path in the scene corresponding to the target video stream, and setting attribute information corresponding to the designated motion path, the attribute information including motion time and motion form; S300, obtaining shooting information of each point on the specified motion path, wherein the shooting information includes a shooting angle and a shooting distance, wherein the shooting angle of a certain point refers to the angle between the line between the point and the corresponding shooting device and the shooting sight line of the shooting device, and the shooting distance of a certain point refers to the distance between the point and the corresponding shooting device; obtaining a shooting information set C corresponding to the motion path = {C 1 , C 2 , ..., C i , ..., C n }, the i-th shooting information C in C i is the shooting information of the i-th point on the specified motion path, C i =(α i , d i ), i ranges from 1 to n, n is the number of points on the specified motion path, α i is the shooting angle corresponding to the i-th point, d i is the shooting distance corresponding to the i-th point; S400, based on the image set to which the specified moving object needs to be added in the target video stream, based on C and the 3D model corresponding to the specified moving object, obtain the two-dimensional image corresponding to each point of the specified moving object, and put the obtained two-dimensional image into the corresponding position of the corresponding image in the target video stream.
2. The method according to claim 1, It is characterized in that S400 specifically includes: S410, set variable counter j=1; S420, if j≤m, execute S430, otherwise, execute S470; m is the number of frames of the image in the image set; S430: for the j-th frame image IMG in the image set, j , get IMG j and the previous frame image IMG j-1 The time interval △t j(j-1) ; S440, based on the starting position, end position, movement time, movement form, IMG corresponding to the specified movement path j-1 The corresponding added point position G j-1 and △t j(j-1) , get IMG j The position G of the adding point where the specified moving object needs to be added j ; G 1 is the starting point position corresponding to the specified motion path; S450, based on G j , obtain the corresponding shooting information from C, and based on G j The corresponding shooting information controls the virtual camera corresponding to the 3D model to shoot the specified moving object in the 3D model, and obtains the corresponding two-dimensional image as G j The corresponding two-dimensional image DIMG j ; S460, from DIMG j The specified moving object is segmented out and the segmented specified moving object is added to IMG j G j Department; S470, adding the specified motion image to the IMG j Add to the current intermediate image set; set j=j+1, and execute S420; the initial value of the current intermediate image set is empty; S480: Use the current intermediate image set as the target image set corresponding to the target video stream.
3. The method according to claim 1, It is characterized in that S400 specifically includes: S411, based on C, generating a camera movement path corresponding to the virtual camera corresponding to the 3D model; S412, based on the camera movement path, controlling the virtual camera to shoot, obtaining a corresponding shooting image set IMGS={DIMG 1 ,DIMG 2 , ..., DIMG i , ..., DIMG m }, DIMG i C based virtual camera i The captured two-dimensional image; S413, set variable counter j=1; S414, if j≤m, execute S415, otherwise, execute S420; m is the number of frames of the image in the image set; S415: for the j-th frame image IMG in the image set, j , get IMG j and the previous frame image IMG j-1 The time interval △t j(j-1) ; S416, based on the starting position, end position, movement time, movement form, IMG corresponding to the specified movement path j-1 The corresponding added point position G j-1 and △t j(j-1) , get IMG j The position G of the adding point where the specified moving object needs to be added j ; G 1 is the starting point position corresponding to the specified motion path; S417, Get from IMGS to IMG j The corresponding two-dimensional image, as IMG j The corresponding two-dimensional image DIMG j ; S418, from DIMG j The specified moving object is segmented out and the segmented specified moving object is added to IMG j G j Department; S419, add the specified motion image to the IMG j Add to the current intermediate image set; set j=j+1, and execute S414; the initial value of the current intermediate image set is empty; S420: Use the current intermediate image set as the target image set corresponding to the target video stream.
4. The method according to claim 2 or 3, It is characterized in that Specify the center of the bounding box of the moving object as the corresponding added point.
5. The method according to claim 2, It is characterized in that S460 and later also include the following steps: S462, for IMG j Identify the target object in the image to obtain a corresponding identification frame; the target object includes the specified moving object; S464, if DIMG j The identification box and IMG corresponding to the specified moving object segmented j There is an overlapping area between any recognition boxes r in d j <d r , block the area in the overlapping area that belongs to the recognition box r. If d j >d r , the overlapping area belongs to DIMG j The area of the recognition box corresponding to the specified moving object segmented from the image is blocked, d r is the distance between the recognition frame r and the corresponding camera device, j ranges from 1 to f(i)-1, and f(i) is IMG j The number of identification boxes in ; execute S470.
6. The method according to claim 3, It is characterized in that S418 and subsequent steps include: S4181, for IMG j Identify the target object in the image to obtain a corresponding identification frame; the target object includes the specified moving object; S4182, if DIMG j The identification box and IMG corresponding to the specified moving object segmented j There is an overlapping area between any recognition boxes r in d j <d r , block the area in the overlapping area that belongs to the recognition box r. If d j >d r , the overlapping area belongs to DIMG j The area of the recognition box corresponding to the specified moving object segmented from the image is blocked, d r is the distance between the recognition frame r and the corresponding camera device, j ranges from 1 to f(i)-1, and f(i) is IMG j The number of identification boxes in ; execute S419.
7. The method according to claim 1, It is characterized in that The designated moving object is a pedestrian or a vehicle.
8. A device for expanding a visual model training data set, It is characterized in that The device comprises: A first acquisition module is used to acquire a video stream to which a specified moving object is currently to be added as a target video stream; A path generation module, used to generate a specified motion path in the scene corresponding to the target video stream, and set attribute information corresponding to the specified motion path, wherein the attribute information includes motion time and motion form; The second acquisition module is used to acquire the shooting information of each point on the specified motion path, wherein the shooting information includes a shooting angle and a shooting distance, wherein the shooting angle of a certain point refers to the angle between the line between the point and the corresponding shooting device and the shooting sight line of the shooting device, and the shooting distance of a certain point refers to the distance between the point and the corresponding shooting device; obtain the shooting information set C corresponding to the motion path = {C 1 , C 2 , ..., C i , ..., C n }, the i-th shooting information C in C i is the shooting information of the i-th point on the specified motion path, C i =(α i , d i ), i ranges from 1 to n, n is the number of points on the specified motion path, α i is the shooting angle corresponding to the i-th point, d i is the shooting distance corresponding to the i-th point; An adding module is used to obtain the two-dimensional image corresponding to each point of the specified moving object based on the image set to which the specified moving object needs to be added in the target video stream, based on C and the 3D model corresponding to the specified moving object, and put the obtained two-dimensional image into the corresponding position of the corresponding image in the target video stream.
9. An electronic device, It is characterized in that including a processor and a memory; The processor is used to execute the steps of the method according to any one of claims 1 to 8 by calling the program or instruction stored in the memory.
10. A non-transitory computer-readable storage medium, It is characterized in that The non-transitory computer-readable storage medium is used to store a program or instruction, and the program or instruction enables a computer to execute the steps of the method according to any one of claims 1 to 8.
Citation Information
Patent Citations
Method and device for generating training samples
CN112308103A
Multi-mode-based object image augmentation method and device, equipment and storage medium
CN114663810A
Motion track generation method and device based on three-dimensional model and server
CN114708381A
Training data set generation method and device, electronic equipment and storage medium
CN117671411A
Three-dimensional reconstruction method, apparatus and system, model training method and storage medium
US20210375036A1
