Data processing method, device, medium, program product and vehicle
By adjusting the multimodal large model and combining it with real-world scenario data and mining prompts, long-tail data in intelligent driving scenarios is automatically collected, solving the problem of low efficiency in manual data collection and improving data collection efficiency and model performance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BYD CO LTD
- Filing Date
- 2025-10-31
- Publication Date
- 2026-07-14
AI Technical Summary
In existing technologies, the method of manually collecting long-tail data is inefficient and difficult to effectively improve the performance of intelligent driving perception models, especially the performance of mining rare long-tail data.
By adjusting the first model, using synthetic road data and real-world scenario data, a multimodal large model is generated and iteratively trained. Combined with the mining of clue words, target road data is automatically collected, improving the model's mining performance.
It enables the automated and efficient extraction of long-tail data from target scenarios from a large amount of real data, improving the collection efficiency of target road data and the generalization ability of the model.
Smart Images
Figure CN122388367A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data mining, and more particularly to a data processing method, device, medium, program product, and vehicle. Background Technology
[0002] Current intelligent driving perception models rely on a large amount of precisely labeled data. The massive amounts of automated road data collected contain a significant amount of long-tail data, which is crucial for improving and optimizing intelligent driving systems. Long-tail data is characterized by its rarity and importance; therefore, mining and labeling this data is essential for enhancing model performance. However, manually collecting long-tail data and other target road data suffers from slow collection efficiency. Summary of the Invention
[0003] This application provides a data processing method, device, medium, program product, and vehicle, which improves the efficiency of target road data collection and at least partially solves the above-mentioned technical problems.
[0004] To achieve the above objectives, according to a first aspect of this application, a data processing method is provided, the method comprising: The first path data is input into the adjusted first model to obtain the target path data belonging to the target scene in the first path data. The first model is then adjusted based on the synthetic path data corresponding to the target scene.
[0005] Optionally, adjusting the first model based on the synthetic road data includes: Based on the synthetic road data and the second road data corresponding to the target scene, hybrid road data is obtained; The first model is adjusted based on the hybrid road sampling data.
[0006] Optionally, the second data source is obtained from the abnormal data fed back by the intelligent driving model corresponding to the vehicle.
[0007] Optionally, adjusting the first model based on the hybrid road sampling data includes: The first model is adjusted based on the hybrid road survey data and the mining prompts associated with the target scene.
[0008] Optionally, the hybrid road survey data includes test road survey data, and the adjustment of the first model based on the hybrid road survey data and the mining prompts associated with the target scene includes: The first model is adjusted based on the training road sampling data and the mining prompts.
[0009] Optionally, adjusting the first model based on the training road sampling data and the mined prompt words includes: The training path data and the mining prompts are input into the first model to obtain the first mining result; The first model is adjusted based on the first mining result, and the training road sampling data and the mining prompt words are re-inputted into the first model based on the adjusted first model to obtain the first mining result, until the first mining result meets the first preset mining condition.
[0010] Optionally, the hybrid road sampling data includes test road sampling data, and after adjusting the first model based on the training road sampling data and the mining prompts, it further includes: The mining prompts are adjusted based on the test road data and the adjusted first model.
[0011] Optionally, adjusting the mining prompts based on the test road data and the adjusted first model includes: The test road data and the mining prompts are input into the adjusted first model to obtain the second mining result; The mining prompts are adjusted based on the second mining result, and the test road data and the mining prompts are re-inputted into the adjusted first model based on the adjusted mining prompts to obtain the second mining result, until the second mining result meets the second preset mining condition.
[0012] Optionally, the step of inputting the first path data into the adjusted first model to obtain the target path data belonging to the target scene from the first path data includes: Input the first road sampling data and the adjusted mining prompts into the adjusted first model to obtain the target road sampling data belonging to the target scene from the first road sampling data.
[0013] Optionally, after inputting the first path data into the adjusted first model to obtain the target path data belonging to the target scene from the first path data, the method further includes: The intelligent driving model is iteratively trained based on the target road data, and then the intelligent driving model is sent to the vehicle.
[0014] Optionally, the method for acquiring the synthetic route data includes: Obtain the real scene data corresponding to the target scene; The real-world scene data is input into the second model to generate data, thus obtaining the synthetic road data corresponding to the target scene.
[0015] Optionally, the method further includes: If the synthetic route data meets the preset requirements, the first model is adjusted based on the synthetic route data.
[0016] Optionally, the preset requirements include that the intelligent driving model iteratively trained based on the synthetic road data meets preset test indicators.
[0017] Optionally, training the intelligent driving model based on the synthetic road data includes: The intelligent driving model is iteratively trained based on the synthetic road data and the third road data.
[0018] Optionally, the method further includes: If the synthetic road data does not meet the preset requirements, the second model is adjusted, and based on the adjusted second model, the real scene data corresponding to the target scene is re-inputted into the second model to generate data, thereby obtaining the synthetic road data corresponding to the target scene.
[0019] Optionally, the real-world scene data includes at least one of real-world scene information, 3D asset information, and sensor parameters.
[0020] According to a second aspect of this application, an electronic device is also provided, including a processor connected to a memory storing a computer program, the processor being configured to run the computer program in the memory to perform any of the data processing methods described above.
[0021] According to a third aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the data processing methods provided in the embodiments of this application.
[0022] According to a fourth aspect of this application, a computer program product is provided, comprising a computer program that is executed by a processor to implement any of the data processing methods provided in the embodiments of this application.
[0023] According to a fifth aspect of this application, a vehicle is provided that performs any of the data processing methods provided in the embodiments of this application, or includes any of the electronic devices provided in the embodiments of this application.
[0024] In summary, in this embodiment of the application, after adjusting the first model based on the synthetic road data corresponding to the target scene, the first road data is input into the adjusted first model to obtain the target road data belonging to the target scene in the first road data. This can automatically and accurately mine the target road data belonging to the target scene from the first road data, thereby improving the collection efficiency of the target road data.
[0025] Other features and advantages of this application will be described in detail in the following detailed description section. Attached Figure Description
[0026] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0027] To gain a more complete understanding of this application and its beneficial effects, the following description will be provided in conjunction with the accompanying drawings, wherein the same reference numerals in the following description denote the same parts.
[0028] Figure 1 This is a flowchart illustrating one embodiment of the data processing method provided in this invention. Figure 2 This is a schematic diagram of the process for generating synthetic route data provided in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the acquisition of the second long-tail data provided in this embodiment of the invention; Figure 4 This is a flowchart illustrating the adjustment model provided in this embodiment of the invention; Figure 5 This is a flowchart illustrating the iterative intelligent model provided in this embodiment of the invention; Figure 6 This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present invention. Detailed Implementation
[0029] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the protection scope of this application.
[0030] Based on the issues mentioned in the background technology, current intelligent driving perception models rely on a large amount of precisely labeled data. The massive amount of automatically collected road data includes a significant amount of long-tail data. In the field of intelligent driving, long-tail data refers to data on traffic scenarios, events, or objects that occur relatively infrequently but are diverse in type. This long-tail data is particularly important for improving and optimizing intelligent driving systems. Long-tail data is characterized by its rarity and importance; therefore, mining and labeling long-tail data is crucial for improving model performance. However, manually collecting long-tail data and other target road data suffers from slow collection efficiency. Automated methods for collecting long-tail data and other target road data also have certain problems. Due to the rarity and limited volume of long-tail data, the performance of the fine-tuned large model in mining long-tail data is relatively low.
[0031] To address the aforementioned issues, this application proposes a data processing method, device, medium, program product, and vehicle. In this application, first road sampling data is input into an adjusted first model to obtain target road sampling data belonging to the target scenario from the first road sampling data. The first model is adjusted based on the synthetic road sampling data corresponding to the target scenario, which can improve the collection efficiency of the target road sampling data.
[0032] Specifically, the data processing method in this application can be applied to electronic devices, which can be cloud-based, vehicle-mounted, or terminal devices. The following description uses an electronic device as an example to illustrate the various embodiments.
[0033] This application provides a data processing method; please refer to [link / reference]. Figure 1 The data processing method provided in this application includes steps S10-S20, which will be described in detail below.
[0034] S10. Input the first path data into the adjusted first model to obtain the target path data belonging to the target scene in the first path data. The first model is adjusted based on the synthetic path data corresponding to the target scene.
[0035] In this embodiment, synthetic road survey data corresponding to the target scene is first acquired to adjust the first model. The synthetic road survey data can be generated by a second model, which has generative capabilities. The second model can be a generative large model, which refers to a deep neural network that learns the joint distribution of data through self-supervised pre-training based on large-scale parameters and massive amounts of data, thereby generating new content according to input prompts. This generation method can generate various types of synthetic road survey data and can effectively adjust the first model to make its mining performance stronger.
[0036] The first model has data mining capabilities. This first large model can be a multimodal large model, which is a large-scale pre-trained model capable of simultaneously receiving, understanding, fusing, and generating multiple data modalities, such as text, images, audio, video, 3D point clouds, and sensor signals. After adjusting the first model using synthetic road data corresponding to the target scene, the first road data is input into the adjusted first model to obtain target road data belonging to the target scene from the first road data. The target scene can be set according to requirements. In this embodiment, it mainly refers to traffic scenes with relatively low frequency but diverse types under intelligent driving scenarios. The target road data is long-tail data in the intelligent driving field.
[0037] In one embodiment, the method for acquiring synthetic route data includes: Obtain the real scene data corresponding to the target scene; The real-world scene data is input into the second model to generate data, thus obtaining the synthetic road data corresponding to the target scene.
[0038] In this embodiment, considering that synthesizing data using a generative large model alone cannot fully simulate the complexity of the real world and has limited generalization ability, real scene data corresponding to the target scene can also be obtained. This real scene data can be real data belonging to the target scene. The real scene data is input into the second large model, and data is generated based on the real scene data to obtain synthetic road data belonging to the target scene. By continuously iterating this step, and after manual quality inspection and accumulation, a large amount of synthetic road data can be generated to adjust the first model. Since real scene data is used as the basis for generating synthetic road data, the generalization ability of the first model can be improved.
[0039] In one embodiment, the method further includes: If the synthetic route data meets the preset requirements, the first model is adjusted based on the synthetic route data.
[0040] In this embodiment, after generating synthetic route data, it is also necessary to determine whether the synthetic route data meets the preset requirements. The preset requirements are that the synthetic route data can be used to adjust the standard of the first model. If the synthetic route data meets the preset requirements, it indicates that the authenticity and validity of the synthetic route data can meet the standard, and the first model can be adjusted based on the synthetic route data.
[0041] In one embodiment, the preset requirements include that the intelligent driving model iteratively trained based on the synthetic road data meets preset test indicators.
[0042] In this embodiment, after generating synthetic road data, the synthetic road data is added to the intelligent driving model for iterative training to obtain the iteratively trained intelligent driving model. The performance of the optimized intelligent driving model is tested using test data to determine whether the optimized intelligent driving model meets the preset test indicators. The preset test indicators are the standards for the performance of the intelligent driving model to reach a certain level. If the intelligent driving model iteratively trained based on the synthetic road data meets the preset test indicators, it indicates that the training of the intelligent driving model with the synthetic road data is effective and the synthetic road data is also real and effective. The first model can be adjusted based on the synthetic road data, which can effectively improve the mining performance of the first model.
[0043] In one embodiment, training the intelligent driving model based on the synthetic road data includes: The intelligent driving model is iteratively trained based on the synthetic road data and the third road data.
[0044] In this embodiment, before iteratively training the intelligent driving model using synthetic road data, the synthetic road data and third-party road data can be mixed together and added to the iterative training of the intelligent driving model. The third-party road data consists of tens of thousands of frames of road data collected daily, which can simulate a more realistic training process, effectively iteratively train the intelligent driving model, improve the generalization ability of the intelligent driving model, and thus more accurately judge the validity of the synthetic road data.
[0045] In one embodiment, the method further includes: If the synthetic road data does not meet the preset requirements, the second model is adjusted, and based on the adjusted second model, the real scene data corresponding to the target scene is re-inputted into the second model to generate data, thereby obtaining the synthetic road data corresponding to the target scene.
[0046] In this embodiment, if the synthetic road data does not meet the preset requirements, the second model needs to be adjusted, the generation strategy of the second model needs to be adjusted, and then based on the adjusted second model, the real scene data corresponding to the target scene is re-inputted into the second model to generate data, so as to obtain new synthetic road data corresponding to the target scene. Then, it is re-determined whether the new synthetic road data meets the preset requirements. Through continuous iteration, manual quality inspection and accumulation, a large amount of synthetic road data can be obtained.
[0047] In one embodiment, the real-world scene data includes at least one of real-world scene information, 3D asset information, and sensor parameters.
[0048] In this embodiment, real-world scene information refers to information that represents a driving scenario. For example, if a tire falls off the road while driving, the information related to this scenario is called real-world scene information. Real-world scene information includes scene design information, semantic information, and annotation rules. Scene design information includes object information within the scene, semantic information describes the categories of each object, and annotation rules are used to associate the mapping relationship between semantic information and scene design information. 3D asset information refers to information generated from 3D modeling; it is a 3D model used to represent objects in the scene. In the target scene, this information is generally obstacle information, such as tires, traffic cones, water barriers, and cars encountered in daily life. Sensor parameters refer to the parameters of sensors on the vehicle, such as the model of the LiDAR, the model of the camera, and the intrinsic and extrinsic parameter matrices. This information can be used to generate small batches of synthetic road data.
[0049] In one example, such as Figure 2 As shown, by combining real-world scene information, 3D asset information, and sensor parameters from the target scenario, a generative large model is used to generate small batches of synthetic road data. These small batch datasets are then mixed with third-party road data and fed into the intelligent driving model for iterative training. The model performance is tested on a test set. If the test metrics meet the requirements, these small batches of data are collected; otherwise, the generative large model is adjusted, and synthetic road data generation continues, iterating continuously. After manual quality control and accumulation, a large amount of synthetic road data can be generated.
[0050] In one embodiment, adjusting the first model based on the synthetic road data includes: Based on the synthetic road data and the second road data corresponding to the target scene, hybrid road data is obtained; The first model is adjusted based on the hybrid road sampling data.
[0051] In this embodiment, the synthesized road data and the second road data corresponding to the target scene collected in reality are mixed to obtain hybrid road data. The second road data is a small amount of road data belonging to the target scene encountered in real driving, which belongs to real long-tail data. The first model is fine-tuned using the hybrid road data. By fine-tuning the first model, the generalization performance of the first model can be increased.
[0052] In one embodiment, the second data source is obtained from abnormal data fed back by the intelligent driving model on the vehicle.
[0053] In this embodiment, the second-path data is obtained from the abnormal data fed back by the intelligent driving model on the vehicle. The intelligent driving model on the vehicle is in actual use and participates in a large number of driving scenarios. When the intelligent driving model feeds back abnormal data, it indicates that the intelligent driving model cannot process the data normally. This data belongs to the long-tail data in the target scenario that is difficult to encounter. After manual screening and collection, this abnormal data can be used as the real long-tail second-path data.
[0054] In one example, such as Figure 3 As shown. During real-vehicle testing, personnel discovered that the vehicle failed to avoid obstacles. They recorded detailed information about the problem and analyzed its causes. For example, a tire that had fallen onto the road was not avoided by the vehicle. The intelligent driving model was unable to identify obstacles such as the tire. However, the model's training set lacked such data. These abnormal data were manually collected as secondary sampling data for subsequent iterations of the intelligent driving model.
[0055] In one embodiment, adjusting the first model based on the hybrid road sampling data includes: The first model is adjusted based on the hybrid road survey data and the mining prompts associated with the target scene.
[0056] In this embodiment, the hybrid road survey data can be used to adjust the first model, and mining prompt words can be written for the required target scenario. Adjusting the first model based on the hybrid road survey data and the mining prompt words associated with the target scenario can improve the generalization ability of the model.
[0057] In one embodiment, the hybrid road sampling data includes test road sampling data, and the adjustment of the first model based on the hybrid road sampling data and the mining prompts associated with the target scene includes: The first model is adjusted based on the training road sampling data and the mining prompts.
[0058] In this embodiment, the hybrid road sampling data can be divided into training road sampling data and test road sampling data. The training road sampling data is the training set, and the test road sampling data is the test set, used to test the fine-tuned first model. Adjusting the first model based on the training road sampling data in the hybrid road sampling data and the mining prompts associated with the target scene can further improve the model's generalization ability.
[0059] In one embodiment, adjusting the first model based on the training road sampling data and the mined prompt words includes: The training path data and the mining prompts are input into the first model to obtain the first mining result; The first model is adjusted based on the first mining result, and the training road sampling data and the mining prompt words are re-inputted into the first model based on the adjusted first model to obtain the first mining result, until the first mining result meets the first preset mining condition.
[0060] In this embodiment, a large amount of non-target scene road data and training road data are mixed and mined with prompt words and input into the first model to obtain the first mining result. The first mining result is the road data mined by the first model from the mixed road data based on the mining prompt words. The road data is compared with the training road data. If the matching degree is higher than the first preset matching degree, it indicates that the first mining result meets the first preset mining condition. Otherwise, the first model needs to be adjusted, and the training road data and mining prompt words are re-inputted into the adjusted first model to obtain a new first mining result. This process is repeated iteratively until the first mining result meets the first preset mining condition, thereby achieving fine-tuning of the first model.
[0061] In one embodiment, the hybrid road sampling data includes test road sampling data, and after adjusting the first model based on the training road sampling data and the mined prompts, the method further includes: The mining prompts are adjusted based on the test road data and the adjusted first model.
[0062] In this embodiment, after adjusting the first model, the mining prompt words will be further adjusted based on the test road data and the adjusted first model, thereby improving the accuracy of the mining prompt words and improving the mining effect of the first model.
[0063] In one embodiment, adjusting the mining prompt words based on the test road data and the adjusted first model includes: The test road data and the mining prompts are input into the adjusted first model to obtain the second mining result; The mining prompts are adjusted based on the second mining result, and the test road data and the mining prompts are re-inputted into the adjusted first model based on the adjusted mining prompts to obtain the second mining result, until the second mining result meets the second preset mining condition.
[0064] In this embodiment, a large amount of non-target scene road data and test road data are mixed and fed into the adjusted first model with mining prompts to obtain a second mining result. The second mining result is the road data mined by the adjusted first model from the mixed road data based on the mining prompts. The road data is compared with the test road data. If the matching degree is higher than the second preset matching degree, it indicates that the second mining result meets the second preset mining condition. Otherwise, the mining prompts need to be adjusted. The training road data and the adjusted mining prompts are re-inputted into the first model to obtain a new second mining result. This process is repeated iteratively until the second mining result meets the second preset mining condition, thereby realizing the adjustment of the mining prompts.
[0065] In one example, such as Figure 4 As shown, synthetic road sampling data is mixed with second-path sampling data corresponding to the real target scene. The multimodal large model is then fine-tuned, and mining prompts are written and input into the fine-tuned multimodal large model. Mining is performed on the test set, and the mining prompts are continuously adjusted based on the mining results until the test indicators meet the requirements. By fine-tuning the multimodal large model, the generalization performance of the model is improved.
[0066] In one embodiment, the step of inputting the first path data into the adjusted first model to obtain the target path data belonging to the target scene from the first path data includes: Input the first road sampling data and the adjusted mining prompts into the adjusted first model to obtain the target road sampling data belonging to the target scene from the first road sampling data.
[0067] In this embodiment, when applying the adjusted first model, a large amount of first-path data to be mined and the adjusted mining prompts can be input into the adjusted first model. The first model vectorizes and extracts the data information from the first-path data and stores the data information. The long-tail data mined by the mining prompts is the target path data belonging to the target scene, which is the target path data belonging to the target scene. The target path data belonging to the target scene in the first-path data can be obtained quickly and accurately.
[0068] In one embodiment, after inputting the first path data into the adjusted first model to obtain the target path data belonging to the target scene from the first path data, the method further includes: The intelligent driving model is iteratively trained based on the target road data to update the corresponding intelligent driving model on the vehicle.
[0069] In this embodiment, the target road data is labeled automatically or manually to generate a large amount of labeled, real training data containing target road data in the target scenario. Then, the intelligent model is iteratively trained based on this training data to optimize the corresponding intelligent driving model on the vehicle.
[0070] It should be noted that the intelligent driving model corresponding to the vehicle can be an online or offline model. It can be deployed on the vehicle or in the cloud and can be used to realize intelligent driving on the vehicle.
[0071] In one example, such as Figure 5 As shown, a large amount of real first-path data is continuously input into the adjusted multimodal large model for vectorization extraction and storage of data information. By inputting adjusted mining prompts, specified long-tail data is mined and then sent into the automatic labeling process to produce a large amount of labeled training data for iterative training of the intelligent driving model.
[0072] In one example, the technical solution provided in this embodiment can solve the problem of serious orientation errors in extra-long vehicles and trucks, achieving beneficial results. Using a finely tuned multimodal large model, more than 3,000 frames of large vehicle data were extracted from over 1 million real-world autonomous driving data points for model iteration, thus resolving the problem of inaccurate large vehicle orientation.
[0073] In one example, abnormal data is fed back from the online intelligent driving model. By analyzing and collecting this long-tail data, second-path data is obtained. Using a generative large model, synthetic road data is generated based on real-world scenario data. The synthetic road data is mixed with the collected real second-path data and the multimodal large model is fine-tuned. Using the fine-tuned multimodal large model, target road data mined from a large amount of real first-path data is used as real long-tail data. This process iterates the online intelligent driving model to achieve a data closed loop.
[0074] Accordingly, embodiments of this application also provide an electronic device, such as... Figure 6 As shown, Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device 1100 further includes a processor 1101 with one or more processing cores, a memory 1102 with one or more computer-readable storage media, and a computer program stored on the memory 1102 and executable on the processor. The processor 1101 and the memory 1102 are electrically connected. Those skilled in the art will understand that the electronic device structure shown in the figure does not constitute a limitation on the electronic device, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.
[0075] The processor 1101 is the control center of the electronic device 1100. It connects various parts of the electronic device 1100 via various interfaces and lines. By running or loading software programs and / or units stored in the memory 1102, and by calling data stored in the memory 1102, it executes various functions and processes data of the electronic device 1100, thereby providing overall monitoring of the electronic device 1100. The processor 1101 can be a processor (Central Processing Unit, CPU), a graphics processing unit (GPU), a network processor (NP), etc., and can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application.
[0076] In this embodiment, the processor 1101 in the electronic device 1100 loads the instructions corresponding to the processes of one or more applications into the memory 1102 according to the following steps, and the processor 1101 runs the applications stored in the memory 1102 to realize various functions, such as: The first path data is input into the adjusted first model to obtain the target path data belonging to the target scene in the first path data. The first model is then adjusted based on the synthetic path data corresponding to the target scene.
[0077] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.
[0078] Optional, such as Figure 6 As shown, the electronic device 1100 also includes: a touch display screen 1103, a radio frequency circuit 1104, an audio circuit 1105, an input unit 1106, and a power supply 1107. The processor 1101 is electrically connected to the touch display screen 1103, the radio frequency circuit 1104, the audio circuit 1105, the input unit 1106, and the power supply 1107. Those skilled in the art will understand that... Figure 6 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.
[0079] The touch display screen 1103 can be used to display a graphical user interface (GUI) and receive operation commands generated by the user interacting with the GUI. The touch display screen 1103 may include a display panel and a touch panel. The display panel can be used to display information input by the user or information provided to the user, as well as various graphical user interfaces of the electronic device. These graphical user interfaces can be composed of graphics, text, icons, video, and any combination thereof. Optionally, the display panel can be configured using a liquid crystal display (LCD), organic light-emitting diode (OLED), or other similar technologies. The touch panel can be used to collect touch operations performed by the user on or near it (such as operations performed by the user using a finger, stylus, or any suitable object or accessory on or near the touch panel), generate corresponding operation commands, and execute the corresponding program according to the operation commands. Optionally, the touch panel may include two parts: a touch detection device and a touch controller. The touch detection device detects the user's touch location and the signal generated by the touch operation, transmitting the signal to the touch controller. The touch controller receives touch information from the touch detection device, converts it into touch point coordinates, and sends it to the processor 1101. It can also receive and execute commands from the processor 1101. The touch panel can cover the display panel. When the touch panel detects a touch operation on or near it, it transmits the information to the processor 1101 to determine the type of touch event. Subsequently, the processor 1101 provides corresponding visual output on the display panel based on the type of touch event. In this embodiment, the touch panel and the display panel can be integrated into the touch display screen 1103 to achieve input and output functions. However, in some embodiments, the touch panel and the touch display screen 1103 can be used as two independent components to achieve input and output functions. That is, the touch display screen 1103 can also be used as part of the input unit 1106 to achieve input functions.
[0080] The radio frequency circuit 1104 can be used to transmit and receive radio frequency signals to establish wireless communication with networked medical devices or other electronic devices, and to transmit and receive signals with networked medical devices or other electronic devices.
[0081] Audio circuit 1105 can be used to provide an audio interface between a user and an electronic device via a speaker and a microphone. Audio circuit 1105 can convert received audio data into electrical signals and transmit them to the speaker, where the speaker converts them into sound signals for output. Conversely, the microphone converts the collected sound signals into electrical signals, which are then received by audio circuit 1105, converted back into audio data, and then processed by processor 1101 before being transmitted via radio frequency circuit 1104 to, for example, another electronic device, or output to memory 1102 for further processing. Audio circuit 1105 may also include an earphone jack to provide communication between peripheral headphones and electronic devices.
[0082] The input unit 1106 can be used to receive input numbers, characters, or user characteristic information (such as fingerprints, iris, facial information, etc.), and to generate keyboard, mouse, joystick, optical, or trackball signal inputs related to user settings and function control.
[0083] Power supply 1107 is used to supply power to various components of electronic device 1100. Optionally, power supply 1107 can be logically connected to processor 1101 through a power management device, thereby enabling functions such as charging, discharging, and power consumption management through the power management device. Power supply 1107 may also include one or more DC or AC power supplies, recharging devices, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.
[0084] although Figure 6 As not shown in the diagram, the electronic device 1100 may also include a camera, sensor, wireless fidelity module, Bluetooth module, etc., which will not be described in detail here.
[0085] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.
[0086] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.
[0087] Therefore, embodiments of this application provide a computer-readable storage medium storing a plurality of computer programs. These computer programs can be loaded by a processor to execute any of the data processing methods provided in this application. The computer program can execute the steps of the following data processing method: The first path data is input into the adjusted first model to obtain the target path data belonging to the target scene in the first path data. The first model is then adjusted based on the synthetic path data corresponding to the target scene.
[0088] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.
[0089] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.
[0090] Since the computer-readable storage medium contains a computer program that can implement any of the data processing methods provided in the embodiments of this application, and can execute any of the data processing methods provided in the embodiments of this application, the effects are detailed in the preceding embodiments and will not be repeated here.
[0091] This application also provides a computer program product that can be loaded by a processor to execute any of the data processing methods provided in this application. Specific implementations of each operation of this data processing method can be found in the preceding embodiments and will not be repeated here.
[0092] Since this computer program can execute any of the data processing methods provided in the embodiments of this application, it can achieve the beneficial effects that any of the data processing methods provided in the embodiments of this application can achieve. Therefore, its beneficial effects are detailed in the preceding embodiments and will not be repeated here.
[0093] This application also provides a vehicle that includes any of the above-described electronic devices, computer-readable storage media, computer program products, or performs any of the above-described methods.
[0094] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.
[0095] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.
[0096] The embodiments, implementation methods, and related technical features of this application can be combined and substituted for each other without conflict.
[0097] The above are merely preferred embodiments of this application and are not intended to limit this application in any way. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of this application without departing from the scope of the technical solution of this application shall still fall within the scope of the technical solution of this application.
Claims
1. A data processing method, characterized in that, The method includes: The first path data is input into the adjusted first model to obtain the target path data belonging to the target scene in the first path data. The first model is then adjusted based on the synthetic path data corresponding to the target scene.
2. The method as described in claim 1, characterized in that, Adjusting the first model based on the synthetic road data includes: Based on the synthetic road data and the second road data corresponding to the target scenario, hybrid road data is obtained. The second road data is obtained from the abnormal data fed back by the intelligent driving model corresponding to the vehicle. The first model is adjusted based on the hybrid road survey data and the mining prompts associated with the target scene.
3. The method as described in claim 2, characterized in that, The hybrid road survey data includes test road survey data. The adjustment of the first model based on the hybrid road survey data and the mining prompts associated with the target scene includes: The first model is adjusted based on the training road sampling data and the mining prompts.
4. The method as described in claim 3, characterized in that, The adjustment of the first model based on the training road sampling data and the mined prompt words includes: The training path data and the mining prompts are input into the first model to obtain the first mining result; The first model is adjusted based on the first mining result, and the training road sampling data and the mining prompt words are re-inputted into the first model based on the adjusted first model to obtain the first mining result, until the first mining result meets the first preset mining condition.
5. The method as described in claim 3, characterized in that, The hybrid road sampling data includes test road sampling data. After adjusting the first model based on the training road sampling data and the mined prompt words, the method further includes: The test road data and the mining prompts are input into the adjusted first model to obtain the second mining result; The mining prompts are adjusted based on the second mining result, and the test road data and the mining prompts are re-inputted into the adjusted first model based on the adjusted mining prompts to obtain the second mining result, until the second mining result meets the second preset mining condition.
6. The method as described in claim 5, characterized in that, The step of inputting the first path data into the adjusted first model to obtain the target path data belonging to the target scene from the first path data includes: Input the first road sampling data and the adjusted mining prompts into the adjusted first model to obtain the target road sampling data belonging to the target scene from the first road sampling data.
7. The method according to any one of claims 1-6, characterized in that, After inputting the first path data into the adjusted first model to obtain the target path data belonging to the target scene from the first path data, the process further includes: The intelligent driving model is iteratively trained based on the target road data, and then the intelligent driving model is sent to the vehicle.
8. The method according to any one of claims 1-6, characterized in that, The method for acquiring the synthetic route data includes: Obtain the real scene data corresponding to the target scene; The real-world scene data is input into the second model to generate data, thus obtaining the synthetic road data corresponding to the target scene.
9. The method as described in claim 8, characterized in that, The method further includes: If the synthetic road data meets the preset requirements, the first model is adjusted based on the synthetic road data. The preset requirements include that the intelligent driving model iteratively trained based on the synthetic road data meets the preset test indicators.
10. The method as described in claim 9, characterized in that, Training the intelligent driving model based on the synthetic road data includes: The intelligent driving model is iteratively trained based on the synthetic road data and the third road data.
11. The method as described in claim 8, characterized in that, The method further includes: If the synthetic road data does not meet the preset requirements, the second model is adjusted, and based on the adjusted second model, the real scene data corresponding to the target scene is re-inputted into the second model to generate data, thereby obtaining the synthetic road data corresponding to the target scene.
12. An electronic device, characterized in that, The device includes a processor connected to a memory storing a computer program, the processor being configured to run the computer program in the memory to perform the data processing method according to any one of claims 1 to 11.
13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the data processing method according to any one of claims 1 to 11.
14. A computer program product, characterized in that, It includes a computer program, which is executed by a processor to implement the data processing method according to any one of claims 1 to 11.
15. A vehicle, characterized in that, The vehicle performs the data processing method as described in any one of claims 1-11, or includes the electronic device as described in claim 12.