Synthetic data scene authenticity evaluation method, electronic equipment and storage medium

Abstract

The invention relates to the technical field of machine learning, in particular to a synthetic data scene authenticity evaluation method, electronic equipment and a storage medium, and aims to solve the technical problem of how to accurately, conveniently and efficiently evaluate the authenticity of a synthetic data scene. In order to achieve the purpose, the occupied raster data of the to-be-evaluated synthetic data scene is input into the trained scene authenticity evaluation neural network model, the scene score of the occupied raster data is obtained, the authenticity of the synthetic data scene is evaluated according to the scene score, image data does not need to be generated, the data size in the rendering process is smaller, and the rendering efficiency is improved. According to the method, the overall process of the evaluation process can be simplified, the evaluation efficiency can be improved, the comparison quantity of the evaluation process is more independent and is not influenced by factors such as the authenticity of virtual assets in the synthetic data scene, and a more real, accurate and effective evaluation result can be obtained, so that the authenticity of the synthetic data scene is effectively verified.

Description

Technical Field

[0001] The present application relates to the technical field of machine learning, and in particular to a method for evaluating the authenticity of a synthetic data scene, an electronic device, and a storage medium. Background Art

[0002] Synthetic data is data created by algorithms or models. Synthetic data can amplify the value of real data and effectively solve the pain points of difficult data collection, expensive labeling, and slow iteration. It can not only support model training based on simulation learning, but also build a new paradigm of training and evaluation through closed-loop simulation. Synthetic data can be applied to multiple fields such as autonomous driving.

[0003] In the process of synthetic data production, it is often necessary to place static scenes and dynamic objects to form some abstract data representing the scene as a whole, which is used for subsequent placement of assets to generate synthetic data. However, there is no better means of verifying the authenticity of synthetic data scenes in the prior art.

[0004] A more conventional verification method is to produce the BEV (Bird's-Eye-View) images corresponding to the synthetic data, and then conduct mixed training of the real data and synthetic data for the cam-based BEV (camera-based BEV) model prediction task, and then evaluate the authenticity of the synthetic data scene by judging whether the performance of the model is improved after adding the synthetic data. However, this method requires the production of image data before evaluation, which takes a long time and consumes a lot of rendering computing power. In addition, the evaluation of scene authenticity through BEV images will be affected by the authenticity of the virtual assets in the scene, resulting in the inability to conduct an independent and rigorous verification process of the control variables.

[0005] Accordingly, the art needs a new authenticity evaluation scheme for synthetic data to solve the above problems. Summary of the invention

[0006] In order to overcome the above-mentioned defects, the present application is proposed to solve or at least partially solve the technical problem of how to accurately, conveniently and efficiently evaluate the authenticity of synthetic data scenes.

[0007] In a first aspect, a method for evaluating the authenticity of a synthetic data scene is provided, the method comprising: Obtaining occupancy grid data of the synthetic data scene to be evaluated; Inputting the occupied grid data of the synthetic data scene to be evaluated into a trained scene authenticity evaluation neural network model to obtain a scene score of the occupied grid data; The authenticity of the synthetic data scene to be evaluated is evaluated according to the scene score.

[0008] In a technical solution of the above-mentioned method for evaluating the authenticity of synthetic data scenes, evaluating the authenticity of the synthetic data scenes to be evaluated according to the scene scores includes: The authenticity of the synthetic data scene to be evaluated is evaluated according to the scene score and a preset score threshold.

[0009] In a technical solution of the above-mentioned method for evaluating the authenticity of synthetic data scenes, evaluating the authenticity of the synthetic data scene to be evaluated according to the scene score and a preset score threshold includes: comparing the scenario score to the score threshold; If the scene score is greater than or equal to the score threshold, it is determined that the synthetic data scene to be evaluated has scene authenticity; If the scene score is less than the score threshold, it is determined that the synthetic data scene to be evaluated does not have scene authenticity.

[0010] In a technical solution of the above-mentioned synthetic data scene authenticity evaluation method, the obtaining of the occupied grid data of the synthetic data scene to be evaluated includes: Acquire scene configuration information of the synthetic data scene to be evaluated, wherein the scene configuration information includes the digital asset type and configuration parameters configured in the synthetic data scene to be evaluated; Obtain the occupied grid data of the synthetic data scene to be evaluated according to the scene configuration information.

[0011] In a technical solution of the above-mentioned method for evaluating the authenticity of synthetic data scenes, the digital asset type includes static digital assets and / or dynamic digital assets, and the configuration parameter includes at least one of a position parameter, a volume parameter, and an angle parameter; Acquiring the scene configuration information of the synthetic data scene to be evaluated includes: The digital asset type and configuration parameters corresponding to the static digital assets and / or dynamic digital assets in the synthetic data scene to be evaluated are obtained.

[0012] In a technical solution of the above-mentioned method for evaluating the authenticity of synthetic data scenes, the occupancy grid of the synthetic data scene to be evaluated is obtained according to the scene configuration information, and the data includes: Dividing the synthetic data scene to be evaluated into grids according to a preset grid size; Determining whether each grid is occupied according to the scene configuration information; The occupied grid data is obtained according to the determination result.

[0013] In a technical solution of the above-mentioned method for evaluating the authenticity of synthetic data scenes, the method further includes training the scene authenticity evaluation neural network model based on the following steps: Acquire training data, wherein the training data includes occupied grid data of a synthetic data scene to be evaluated and occupied grid data of a real data scene; The occupied grid data of the synthetic data scene to be evaluated and the occupied grid data of the real data scene are used to train the scene authenticity evaluation neural network model to obtain a trained scene authenticity evaluation neural network model.

[0014] In one technical solution of the above-mentioned synthetic data scene authenticity evaluation method, the scene authenticity evaluation neural network model is a generative adversarial network model.

[0015] In a second aspect, an electronic device is provided, comprising at least one processor; and a memory communicatively connected to the at least one processor; wherein a computer program is stored in the memory, and when the computer program is executed by the at least one processor, the method described in any one of the technical solutions of the above-mentioned synthetic data scene authenticity assessment method is implemented.

[0016] In a third aspect, a computer-readable storage medium is provided, wherein a plurality of program codes are stored therein, wherein the program codes are suitable for being loaded and run by a processor to execute the method described in any one of the technical solutions of the above-mentioned synthetic data scene authenticity assessment method.

[0017] The above one or more technical solutions of the present application have at least one or more of the following beneficial effects: In implementing the technical solution of the synthetic data scene authenticity evaluation method provided by the present application, the present application inputs the occupied grid data of the synthetic data scene to be evaluated into the trained scene authenticity evaluation neural network model, obtains the scene score of the occupied grid data, and evaluates the authenticity of the synthetic data scene according to the scene score. Through the above configuration, the present application can realize the evaluation of the authenticity of the synthetic data scene based on the occupied grid data of the synthetic data scene, without the need to generate image data, and the amount of data in the rendering process is less, which can simplify the overall process of the evaluation process and improve the evaluation efficiency. At the same time, since the occupied grid data only indicates whether the grid is occupied, the comparison amount of the evaluation process is more independent, and will not be affected by factors such as the authenticity of virtual assets in the synthetic data scene, and can obtain more real, accurate and effective evaluation results, thereby achieving effective verification of the authenticity of the synthetic data scene. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The disclosure of the present application will become easier to understand with reference to the accompanying drawings. It is easy for those skilled in the art to understand that these drawings are only for illustrative purposes and are not intended to limit the scope of protection of the present application. Among them: Figure 1 It is a flowchart of the main steps of a method for evaluating the authenticity of a synthetic data scene according to an embodiment of the present application; Figure 2 It is a flowchart of the main steps of a method for evaluating the authenticity of a synthetic data scene according to an implementation of an embodiment of the present application; Figure 3 It is a schematic diagram of the main structure of an electronic device according to an embodiment of the present application.

[0019] Reference numerals: 11: memory; 12: processor. DETAILED DESCRIPTION

[0020] Some embodiments of the present application are 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 application and are not intended to limit the protection scope of the present application.

[0021] In the description of the present application, "module" and "processor" may include hardware, software or a combination of the two. A module may include hardware circuits, various suitable sensors, communication ports, memory, and may also include software parts, such as program code, or a combination of software and hardware. The processor may be a central processing unit, a microprocessor, an image processor, a digital signal processor or any other suitable processor. The processor has data and / or signal processing functions. The processor may be implemented in software, hardware or a combination of the two. Computer-readable storage media include any suitable medium that can store program code, such as a disk, a hard disk, an optical disk, a flash memory, a read-only memory, a random access memory, etc. The term "A and / or B" means all possible combinations of A and B, such as only A, only B or A and B. The term "at least one A or B" or "at least one of A and B" has a similar meaning to "A and / or B" and may include only A, only B or A and B. The singular terms "one" and "the" may also include plural forms.

[0022] The relevant user personal information that may be involved in the various embodiments of this application is strictly in accordance with the requirements of laws and regulations, following the principles of legality, legitimacy and necessity, based on the reasonable purposes of business scenarios, to process the personal information that users actively provide during the use of products / services or generated due to the use of products / services, as well as the personal information obtained with the user's authorization.

[0023] The user personal information processed by this application will vary depending on the specific product / service scenario, and shall be based on the specific scenario in which the user uses the product / service, and may involve the user's account information, device information, driving information, vehicle information or other related information. This application will treat the user's personal information and its processing with a high degree of diligence.

[0024] This application attaches great importance to the security of user personal information and has taken reasonable and feasible security protection measures that meet industry standards to protect user information and prevent personal information from being accessed, disclosed, used, modified, damaged or lost without authorization.

[0025] Here we first explain some terms involved in this application.

[0026] Occupancy grid: A grid map used to represent obstacles and free space in an environment. It divides the space into a series of small grid cells (grids), each of which can be marked as "occupied" or "unoccupied", indicating whether the area is occupied by an obstacle.

[0027] BEV model: Bird's Eye View model is a model that converts environmental data into a bird's-eye view representation, which is usually used for perception and decision-making. The BEV model projects the three-dimensional point cloud or image data obtained by sensors (such as lidar, camera) onto a plane to provide a clear view of the vehicle's surroundings.

[0028] Scene: In the synthetic data production line, some abstract data consisting of static scenes + dynamic targets are placed, which are used for subsequent placement of assets to generate synthetic data.

[0029] See attached Figure 1 , Figure 1 FIG. 1 is a flow chart showing the main steps of a method for evaluating the authenticity of a synthetic data scene according to an embodiment of the present application. Figure 1 As shown, the synthetic data scene authenticity evaluation method in the embodiment of the present application mainly includes the following steps S101 to S103.

[0030] Step S101: Obtaining occupancy grid data of a synthetic data scene to be evaluated.

[0031] In this embodiment, the occupancy grid data of the synthetic data scene to be evaluated may be obtained.

[0032] In one implementation, step S101 may further include the following steps S1011 and S1012: Step S1011: Acquire scene configuration information of the synthetic data scene to be evaluated, where the scene configuration information includes the digital asset type and configuration parameters configured in the synthetic data scene to be evaluated.

[0033] In this embodiment, the digital asset type may include static digital assets and dynamic digital assets, and the configuration parameters may include at least one of position parameters, volume parameters, and angle parameters. The asset types and configuration parameters corresponding to the static digital assets and dynamic digital assets in the synthetic data scene to be evaluated may be obtained as scene configuration information.

[0034] Static digital assets may include buildings, overpasses, trees, cones, etc. Dynamic digital assets may include vehicles, pedestrians, non-motor vehicles, etc. Configuration parameters may include parameter information such as the position, size, angle, etc. of static digital assets and dynamic digital assets.

[0035] Step S1012: Obtaining the occupancy grid data of the synthetic data scene to be evaluated according to the scene configuration information.

[0036] In this embodiment, step S1012 may further include the following steps S10121 to S10123: Step S10121: Divide the synthetic data scene to be evaluated into grids according to a preset grid size.

[0037] In this embodiment, the synthetic data scene to be evaluated can be divided according to a preset grid size. For example, the range of the synthetic data scene to be evaluated is an area of ​​50m in length, 50m in width, and 1.5m in height; the grid size is 200x200x16mm; the area corresponding to the synthetic data scene to be evaluated can be grid-divided according to the grid size.

[0038] Step S10122: Determine whether each grid is occupied according to the scene configuration information.

[0039] In this embodiment, it is possible to determine whether each grid is occupied based on the scene configuration information, that is, to determine whether each grid is occupied based on the position, size, angle and other information of the dynamic digital assets and the static digital assets.

[0040] Step S10123: Obtain occupied grid data according to the judgment result.

[0041] In this embodiment, the occupied grid data of the synthetic data scene to be evaluated may be generated according to the judgment result of whether each grid is occupied.

[0042] Step S102: inputting the occupied grid data of the synthetic data scene to be evaluated into the trained scene authenticity evaluation neural network model to obtain the scene score of the occupied grid data.

[0043] In this embodiment, based on the scene authenticity evaluation neural network model, the occupied grid data of the synthetic data scene to be evaluated can be input into the trained scene authenticity evaluation neural network model to obtain the scene score of the occupied grid data.

[0044] In one implementation, the scene reality neural network model may be a classification model, such as a GAN (Generative Adversarial Network) model.

[0045] In one implementation, the scene reality neural network model may be trained according to the following steps S201 to S202: Step S201: Acquire training data, where the training data includes occupied grid data of a synthetic data scene to be evaluated and occupied grid data of a real data scene.

[0046] Step S202: Using the occupied grid data of the synthetic data scene to be evaluated and the occupied grid data of the real data scene, the scene authenticity evaluation neural network model is trained to obtain a trained scene authenticity evaluation neural network model.

[0047] In this embodiment, the occupied grid data of the synthetic data scene to be evaluated and the occupied grid data of the real data scene can be used to train the scene authenticity evaluation neural network model, so as to obtain a trained scene authenticity evaluation neural network model. The synthetic data scene to be evaluated is evaluated based on the trained scene authenticity evaluation neural network model, so as to obtain the scene score of the synthetic data scene to be evaluated.

[0048] Taking the GAN model as an example of the scene authenticity evaluation neural network model, the occupancy grid data of the synthetic data scene to be evaluated and the occupancy grid data of the real data scene can be used to train the GAN model, thereby obtaining a trained GAN model. The discriminator in the trained GAN model is used to score the synthetic data scene to be evaluated to obtain a scene score. The higher the scene score, the closer the synthetic data scene to be evaluated is to the real data scene.

[0049] In another embodiment, the occupancy grid data of the synthetic data scene used for training may be annotated with a scene score, and the scene authenticity evaluation neural network model may be trained based on the occupancy grid data annotated with the scene score, thereby obtaining a trained scene authenticity evaluation neural network model. The synthetic data scene to be evaluated is evaluated based on the trained scene authenticity evaluation neural network model, thereby obtaining a scene score of the synthetic data scene to be evaluated.

[0050] Step S103: Evaluate the authenticity of the synthetic data scene to be evaluated based on the scene score.

[0051] In this embodiment, the authenticity of the synthetic data scene to be evaluated may be evaluated based on the scene score and a preset score threshold.

[0052] Specifically, the scene score can be compared with a preset score threshold; if the scene score is greater than or equal to the score threshold, it can be determined that the synthetic data scene to be evaluated has scene authenticity; if the scene score is less than the score threshold, it can be determined that the synthetic data scene to be evaluated does not have scene authenticity. Among them, those skilled in the art can set the value of the score threshold according to the needs of actual applications.

[0053] Based on the method described in the above steps S101 to S103, the embodiment of the present application inputs the occupied grid data of the synthetic data scene to be evaluated into the trained scene authenticity evaluation neural network model, obtains the scene score of the occupied grid data, and evaluates the authenticity of the synthetic data scene according to the scene score. Through the above configuration, the embodiment of the present application can realize the evaluation of the authenticity of the synthetic data scene based on the occupied grid data of the synthetic data scene, without the need to generate image data, and the amount of data in the rendering process is less, which can simplify the overall process of the evaluation process and improve the evaluation efficiency. At the same time, since the occupied grid data only indicates whether the grid is occupied, the comparison amount of the evaluation process is more independent, and will not be affected by factors such as the authenticity of virtual assets in the synthetic data scene, and can obtain more real, accurate and effective evaluation results, thereby achieving effective verification of the authenticity of the synthetic data scene.

[0054] In one embodiment, reference may be made to the attached Figure 2 , Figure 2 FIG. 1 is a flow chart of the main steps of a method for evaluating the authenticity of a synthetic data scene according to an implementation of an embodiment of the present application. Figure 2 As shown, the synthetic data scene authenticity evaluation method may include the following steps S301 to S305: Step S301: Acquire scene configuration information, including static digital assets and dynamic digital assets.

[0055] Step S302: Generate occupancy grid data of the synthetic data scene.

[0056] Step S303: Train the GAN model using the occupancy grid data of the synthetic data scene and the occupancy grid data of the real data scene.

[0057] Step S304: Using the trained GAN model, output the scene score of the synthetic data scene.

[0058] Step S305: Determine the scene authenticity of the synthetic data scene according to the scene score.

[0059] It should be pointed out that although the various steps in the above embodiments are described in a specific order, those skilled in the art can understand that in order to achieve the effect of the present application, different steps do not have to be executed in such an order, and they can be executed simultaneously (in parallel) or in other orders. These adjusted schemes are equivalent to the technical schemes described in this application, and therefore will also fall within the scope of protection of this application.

[0060] It is understood by those skilled in the art that all or part of the processes in the method for implementing the above-mentioned embodiment of the present application can also be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable storage medium may include: any entity or device, medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal and software distribution medium, etc. that can carry the computer program code.

[0061] Another aspect of the present application also provides a computer-readable storage medium.

[0062] In an embodiment of a computer-readable storage medium according to the present application, the computer-readable storage medium may be configured to store a program for executing the synthetic data scene authenticity evaluation method of the above method embodiment, and the program may be loaded and run by a processor to implement the above synthetic data scene authenticity evaluation method. For ease of explanation, only the parts related to the embodiment of the present application are shown. For specific technical details not disclosed, please refer to the method part of the embodiment of the present application. The computer-readable storage medium may be a storage device formed by various electronic devices. Optionally, the computer-readable storage medium in the embodiment of the present application is a non-transitory computer-readable storage medium.

[0063] Another aspect of the present application provides an electronic device.

[0064] In an embodiment of an electronic device according to the present application, the electronic device may include at least one processor; and a memory connected to the at least one processor; wherein a computer program is stored in the memory, and when the computer program is executed by the at least one processor, the method described in any of the above embodiments is implemented. The electronic device described in the present application may be, but is not limited to, a mobile phone, a tablet computer, a desktop, a laptop, a handheld computer, a notebook computer, a vehicle-mounted device, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), an augmented reality (AR) device, a virtual reality (VR) device, etc., and the embodiments of the present application are not limited to this. See the attached Figure 3 , Figure 3 FIG. 4 exemplarily shows that the memory 11 and the processor 12 are communicatively connected via a bus.

[0065] So far, the technical solution of the present application has been described in conjunction with an embodiment shown in the accompanying drawings, but it is easy for those skilled in the art to understand that the protection scope of the present application is obviously not limited to these specific embodiments. Without departing from the principles of the present application, 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 application.

Claims

1. A method for evaluating the authenticity of synthetic data scenes, characterized in that: The method comprises: Obtaining occupancy grid data of the synthetic data scene to be evaluated; Inputting the occupied grid data of the synthetic data scene to be evaluated into a trained scene authenticity evaluation neural network model to obtain a scene score of the occupied grid data; The authenticity of the synthetic data scene to be evaluated is evaluated according to the scene score.

2. The method for evaluating the authenticity of synthetic data scenes according to claim 1, characterized in that: The step of evaluating the authenticity of the synthetic data scene to be evaluated according to the scene score includes: The authenticity of the synthetic data scene to be evaluated is evaluated according to the scene score and a preset score threshold.

3. The method for evaluating the authenticity of synthetic data scenes according to claim 2, characterized in that: The step of evaluating the authenticity of the synthetic data scene to be evaluated according to the scene score and a preset score threshold includes: comparing the scenario score to the score threshold; If the scene score is greater than or equal to the score threshold, it is determined that the synthetic data scene to be evaluated has scene authenticity; If the scene score is less than the score threshold, it is determined that the synthetic data scene to be evaluated does not have scene authenticity.

4. The method for evaluating the authenticity of synthetic data scenes according to claim 1, characterized in that: The obtaining of the occupied grid data of the synthetic data scene to be evaluated comprises: Acquire scene configuration information of the synthetic data scene to be evaluated, wherein the scene configuration information includes the digital asset type and configuration parameters configured in the synthetic data scene to be evaluated; Obtain the occupied grid data of the synthetic data scene to be evaluated according to the scene configuration information.

5. The method for evaluating the authenticity of synthetic data scenes according to claim 4, wherein the digital asset type comprises static digital assets and / or dynamic digital assets, and the configuration parameters comprise at least one of position parameters, volume parameters, and angle parameters; characterized in that: Acquiring the scene configuration information of the synthetic data scene to be evaluated includes: The digital asset type and configuration parameters corresponding to the static digital assets and / or dynamic digital assets in the synthetic data scene to be evaluated are obtained.

6. The method for evaluating the authenticity of synthetic data scenes according to claim 4, characterized in that: According to the scene configuration information, the occupied grid of the synthetic data scene to be evaluated is obtained, and the data includes: Dividing the synthetic data scene to be evaluated into grids according to a preset grid size; Determining whether each grid is occupied according to the scene configuration information; The occupied grid data is obtained according to the determination result.

7. The method for evaluating the authenticity of synthetic data scenes according to any one of claims 1 to 6, characterized in that: The method further comprises training the scene authenticity assessment neural network model based on the following steps: Acquire training data, wherein the training data includes occupied grid data of a synthetic data scene to be evaluated and occupied grid data of a real data scene; The occupied grid data of the synthetic data scene to be evaluated and the occupied grid data of the real data scene are used to train the scene authenticity evaluation neural network model to obtain a trained scene authenticity evaluation neural network model.

8. The method for evaluating the authenticity of synthetic data scenes according to claim 7, characterized in that: The scene authenticity evaluation neural network model is a generative adversarial network model.

9. An electronic device, characterized in that: include: at least one processor; and, a memory communicatively coupled to the at least one processor; Wherein, a computer program is stored in the memory, and when the computer program is executed by the at least one processor, the synthetic data scene authenticity assessment method described in any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium storing a plurality of program codes, characterized in that: The program code is suitable for being loaded and run by a processor to execute the synthetic data scene authenticity assessment method according to any one of claims 1 to 8.

Images