Method, device, electronic device and storage medium for generating training data

Abstract

A method, device, electronic device and storage medium for generating data to be trained in an embodiment of the present invention obtains driving incremental data and scenario application types from the vehicle side, uploads the driving incremental data to the cloud, and then classifies the driving incremental data based on the scenario application type to obtain multiple scenario type data sets, and extracts metadata to obtain driving scene metadata, and then converts the driving scene metadata into a data format according to a pre-configured data processing node, and generates data to be trained for the vehicle-side driving model based on the format-converted driving scene metadata; this solution effectively avoids the waste of local storage resources by intermediate data by uploading the data to the cloud and performing corresponding data operations, and effectively improves the operational efficiency of data integrity control based on the metadata extraction of the scene data and the overall labeling and control of the data based on the metadata, thereby facilitating the generation of training data required for the vehicle-side model.

Description

Technical Field

[0001] The present application relates to the field of computers, and in particular to a method, device, electronic device, and storage medium for generating training data. Background Art

[0002] In the process of autonomous driving research and development, vehicle perception, driving decisions, and decision execution are necessary conditions for obtaining high-level autonomous driving capabilities. As perception technology and computing platforms become increasingly mature and convergent, autonomous driving is a science that imitates cognitive logic. Based on the premise of allowing vehicles to perceive road conditions and driving environments, building accurate data analysis models is a prerequisite for cognition. In order to improve cognitive accuracy, it is necessary to continuously collect, analyze, and process more data from new scenarios for model feeding and training, so as to give full play to the advantages of the algorithm and optimize the autonomous driving model.

[0003] However, in related technologies, in the process of processing the amount of autonomous driving data required for model training, it is necessary to perform overall data processing operations such as storage, classification, and data cleaning on the data. Under such data operation status, a large amount of storage space and a long processing cycle are required, which cannot meet the requirements of high-efficiency data integrity operations. Moreover, since the data processing scenarios are distributed on the local storage end and the local control end, a large amount of data storage space is often generated during the processing process to store intermediate data, resulting in a waste of local storage resources. Summary of the Invention

[0004] The purpose of the embodiments of the present invention is to provide a method, device, electronic device and storage medium for generating training data to solve the problem that the existing technology cannot meet the requirements of high-efficiency data integrity operation and causes waste of local storage resources.

[0005] The present invention provides a method for generating data to be trained, which includes: obtaining driving incremental data and scenario application types on the vehicle side, and uploading the driving incremental data to the cloud; classifying the driving incremental data based on the scenario application type to obtain multiple scenario type data sets, and the scenario type data sets include multiple driving incremental data; extracting metadata from the multiple scenario type data groups to obtain driving scene metadata; performing data format conversion on the driving scene metadata according to a pre-configured data processing node, and generating data to be trained for a vehicle-side driving model based on the format-converted driving scene metadata.

[0006] In one embodiment of the present invention, obtaining driving incremental data on the vehicle side includes: obtaining the initial driving data volume on the vehicle side, and monitoring the total data volume of the initial driving data volume on the vehicle side; if the total data volume increases, extracting the initial incremental data, and determining the structure type of the initial incremental data, the structure type including unstructured data and structured parameters; if the incremental data is unstructured data, extracting driving parameters from the unstructured data according to a preset data storage format to obtain vehicle-side incremental parameters; determining the vehicle-side incremental parameters and structured data as the driving incremental data, and storing them in a preset incremental data storage area.

[0007] In one embodiment of the present invention, after the vehicle-side incremental parameter is determined as the driving incremental data, the method for generating data to be trained further includes: obtaining the data collection time of the driving incremental data, and timestamping the driving incremental data according to the data collection time to obtain driving incremental data with a timestamp; generating a storage time series according to a preset time interval, and performing data alignment on the driving incremental data with a timestamp within the time series according to the timestamp, so as to store the driving incremental data according to the time series.

[0008] In one embodiment of the present invention, performing data format conversion on the driving scene metadata according to a pre-configured data processing node includes: determining data dependency and data index relationship of the driving scene metadata based on the topological association relationship of the preset driving module on the vehicle side; generating data dependency labels and data index labels according to the data dependency and data index relationship, and labeling the driving scene metadata according to the data dependency labels and data index labels to obtain labeled driving scene metadata; performing data format conversion on the labeled driving scene metadata according to the preset training data format of the driving model to obtain format-converted driving scene metadata, so as to generate training data for the vehicle-side driving model based on the format-converted driving scene metadata.

[0009] In one embodiment of the present invention, before determining the data dependency and data index relationship of the driving scene metadata based on the topological association relationship of the preset driving module on the vehicle side, the method for generating data to be trained also includes: performing data cleaning processing on the driving scene metadata, and the data cleaning processing includes data deduplication, data fitting and data desensitization to eliminate invalid data from the driving scene metadata.

[0010] In one embodiment of the present invention, generating the training data required for the vehicle-side driving model based on the driving scene metadata after format conversion includes: dividing the driving scene metadata after format conversion into a training data set and a verification data set according to a preset division ratio, and storing the training data set and the verification data set in a preset training data storage area.

[0011] In one embodiment of the present invention, after metadata is extracted from the multiple scene type data groups, the method for generating data to be trained further includes: obtaining data scene key features of the vehicle-side driving model; and performing data retrieval in the extracted initial metadata based on the data scene key features to determine the driving scene metadata.

[0012] An embodiment of the present invention also provides a device for generating data to be trained, which includes: a vehicle-side data acquisition module for acquiring driving incremental data and scene application types on the vehicle side; an incremental data classification module for classifying the driving incremental data based on the scene application type to obtain multiple scene type data sets, and the scene type data sets include multiple driving incremental data; a metadata extraction module for extracting metadata from the multiple scene type data groups to obtain driving scene metadata; a metadata processing module for converting the data format of the driving scene metadata according to a pre-configured data processing node, and generating data to be trained for the vehicle-side driving model based on the driving scene metadata after the format conversion.

[0013] An embodiment of the present invention also provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the electronic device implements the method for generating data to be trained as described in any one of the above embodiments.

[0014] An embodiment of the present invention further provides a computer-readable storage medium having computer-readable instructions stored thereon. When the computer-readable instructions are executed by a processor of a computer, the computer is caused to execute the method for generating data to be trained as described in any one of the above embodiments.

[0015] A method, device, electronic device and storage medium for generating data to be trained in an embodiment of the present invention obtains driving incremental data and scenario application types from the vehicle side, uploads the driving incremental data to the cloud, and then classifies the driving incremental data based on the scenario application type to obtain multiple scenario type data sets, and extracts metadata to obtain driving scene metadata, and then converts the driving scene metadata into a data format according to a pre-configured data processing node, and generates data to be trained for the vehicle-side driving model based on the format-converted driving scene metadata; this solution effectively avoids the waste of local storage resources by intermediate data by uploading data to the cloud and performing corresponding data operations, and effectively improves the operational efficiency of data integrity control based on metadata extraction of scene data and overall labeling and control of data based on metadata, thereby facilitating the generation of training data required for the vehicle-side model.

[0016] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings are incorporated into and constitute a part of the specification, illustrating embodiments consistent with the present application and, together with the specification, serving to explain the principles of the present application. It is obvious that the drawings described below are merely some embodiments of the present application, and a person of ordinary skill in the art can derive other drawings based on these drawings without inventive effort. In the drawings:

[0018] Figure 1 is a schematic diagram of an exemplary system architecture shown in an exemplary embodiment of the present application;

[0019] Figure 2 is a flowchart of a method for generating data to be trained, shown in an exemplary embodiment of the present application;

[0020] Figure 3 is a schematic diagram of a driving incremental data acquisition process shown in an exemplary embodiment of the present application;

[0021] Figure 4 This is a flowchart of a specific method for generating data to be trained, as shown in an exemplary embodiment of the present application;

[0022] Figure 5 This is a flowchart of a specific driving incremental data acquisition shown in an exemplary embodiment of the present application;

[0023] Figure 6 is a schematic diagram of a device for generating data to be trained, shown in an exemplary embodiment of the present application;

[0024] Figure 7It is a structural diagram of a computer system of an electronic device shown in an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0025] The following will describe embodiments of the present invention with reference to the accompanying drawings and specific embodiments. Those skilled in the art will readily appreciate the other advantages and benefits of the present invention from the disclosure herein. The present invention may also be implemented or applied through various other specific embodiments, and the details in this specification may be modified or altered based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are intended only to illustrate the present invention and are not intended to limit the scope of protection of the present invention.

[0026] It should be noted that the illustrations provided in the following embodiments are merely schematic illustrations of the basic concept of the present invention. Therefore, the illustrations only show components related to the present invention and are not drawn according to the number, shape, and size of components in actual implementation. In actual implementation, the type, quantity, and proportion of each component may be changed arbitrarily, and the component layout may also be more complex.

[0027] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the embodiments of the present invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring the embodiments of the present invention.

[0028] The term "and / or" used in this application describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. The character " / " generally indicates that the associated objects are in an "or" relationship.

[0029] First, it's important to note that metadata refers to data that describes data. It provides information about the data, such as its content, format, source, creation date, author, owner, modification history, and relationships. It helps users better understand and use the data. Metadata helps data managers categorize, organize, index, retrieve, and share data, while also helping data users quickly locate and use the data they need. Metadata can encompass a variety of data formats, such as text, images, audio, and video.

[0030] Structured data refers to data that can be organized into tables or similar structures. This data typically has a well-defined format and fields, making it easy to organize, manage, query, and analyze. Common examples of structured data include numbers, dates, times, currencies, prices, names, addresses, and more, and are typically stored in relational databases.

[0031] Unstructured data refers to data that cannot be easily placed into a table or similar structure. This data typically lacks a clear format and may be text, images, audio, video, or a mixture of various formats and fields. Unstructured data is often difficult to manage, organize, query, and analyze, requiring specialized techniques and tools. Common examples of unstructured data include emails, social media posts, news articles, books, blogs, images, and videos, and are often stored in document management systems, content management systems, and big data platforms.

[0032] The present application also provides the following beneficial effects: monitoring the initial amount of vehicle-side driving data, determining data increments, and extracting data from the data increments based on the data structure to obtain data content in a unified and standardized storage format, which facilitates further data processing operations and the retrieval and viewing of intermediate data, and also allows for unified storage; generating a timestamp based on the data acquisition time, annotating the data, and aligning the data based on a unified time series for storage, so that data storage can be arranged based on a unified sequence, facilitating the extraction and processing of data at specific time nodes; determining data dependencies and index relationships based on the topological relationships of the vehicle-side driving modules, which ensures the data's lineage after indexing and dependency labeling, and facilitates the reflection of topological relationships and the determination of the relevance of data structures when retrieving data; deduplication, data fitting, and data desensitization of metadata can ensure the validity of the data, facilitate further processing of the training data, and ensure the accuracy of the data to be trained; retrieving data based on the key features of the data scenario can narrow the data requirements based on the training model and reduce the amount of data to be processed, thereby improving data processing efficiency.

[0033] Figure 1 It is a schematic diagram of an exemplary system architecture shown in an exemplary embodiment of the present application.

[0034] Reference Figure 1As shown, the system architecture may include a vehicle-side 101 and a computer device 102. The vehicle-side 101 is used to obtain the vehicle-side incremental driving data and scenario application types, and upload the incremental driving data to the cloud and provide it to the computer device 102 for processing. The computer device 102 may be at least one of a microcomputer, an embedded computer, a network computer, and the like. Relevant technicians may implement classification of incremental driving data based on scenario application types on the computer device 102 to obtain multiple scenario type data sets, perform metadata extraction on the multiple scenario type data sets to obtain driving scenario metadata, perform data format conversion on the driving scenario metadata according to pre-configured data processing nodes, and generate data to be trained for the vehicle-side driving model based on the format-converted driving scenario metadata.

[0035] Schematically, after the computer device 102 obtains the driving incremental data and scenario application type of the vehicle side 101 and uploads the driving incremental data to the cloud, it classifies the driving incremental data based on the scenario application type to obtain multiple scenario type data sets, and extracts metadata to obtain driving scene metadata. Then, according to the pre-configured data processing node, the driving scene metadata is converted into a data format and then, based on the format-converted driving scene metadata, data to be trained for the vehicle-side driving model is generated. This solution can effectively avoid the waste of local storage resources caused by intermediate data by uploading the data to the cloud and performing corresponding data operations. Moreover, based on the metadata extraction of the scene data and the overall labeling and control of the data based on the metadata, the operational efficiency of the overall data control is effectively improved, which facilitates the generation of the training data required for the vehicle-side model.

[0036] Figure 2 is a flowchart of a method for generating data to be trained, which is shown in an exemplary embodiment of the present application. The method for generating data to be trained can be executed by a computing processing device, which can be Figure 1 The computer device 102 shown in FIG. Figure 2 As shown, the flowchart of the method for generating training data includes at least steps S210 to S240, which are described in detail as follows:

[0037] In step S210, the driving incremental data and scenario application type on the vehicle side are obtained, and the driving incremental data is uploaded to the cloud.

[0038] In one embodiment of the present application, Figure 3 This is a schematic diagram of the driving incremental data acquisition process shown in an exemplary embodiment of the present application. The driving incremental data acquisition process can be Figure 3 The driving incremental data acquisition process is executed as shown in the schematic diagram, Figure 3The driving incremental data acquisition process shown includes at least steps S310 to S340, which are described in detail as follows:

[0039] In step S310, the initial driving data volume of the vehicle side is obtained, and the total data volume of the initial driving data volume of the vehicle side is monitored.

[0040] In one embodiment of the present application, the monitoring process for the total amount of initial driving data from the vehicle can be configured with a monitoring period. This allows for periodic monitoring of data changes to promptly identify data update times. The above-mentioned monitoring period can be adjusted and set based on the accuracy requirements for data monitoring during actual implementation of this solution, and no specific limitation is imposed on the monitoring period.

[0041] In step S320, if the total amount of data increases, the initial incremental data is extracted and the structure type of the initial incremental data is determined.

[0042] In one embodiment of the present application, the above-mentioned structure types include unstructured data and structured parameters.

[0043] In step S330, if the incremental data is unstructured data, driving parameters are extracted from the unstructured data according to a preset data storage format to obtain vehicle-side incremental parameters.

[0044] In one embodiment of the present application, if the incremental data is structured data, a data storage format check is performed on the structured data, and missing values ​​are filled with data fitting, thereby meeting the data storage format consistency requirement and ensuring data validity.

[0045] In step S340, the vehicle-side incremental parameters and structured data are determined as driving incremental data and stored in a preset incremental data storage area.

[0046] In one embodiment of the present application, the above-mentioned preset incremental data storage area is set in a local storage area, and is uploaded to the cloud for processing after the driving incremental data storage is completed.

[0047] The above-mentioned driving incremental data acquisition method monitors the initial driving data volume on the vehicle side, determines the data increment, and extracts the data increment according to the data structure to obtain data content in a unified and standardized storage format, which facilitates further processing operations on the data and the retrieval and viewing of intermediate data, and can also be stored in a unified manner.

[0048] In one embodiment of the present application, a timestamp refers to a number representing a certain point in time, usually an integer or a floating point number. In computer science, a timestamp is usually the time elapsed from a certain fixed point in time, in seconds, milliseconds or microseconds. A timestamp can be used to mark the time when an event occurs, such as the time an email is sent, the time a file is created, the time a web page is accessed, etc. Timestamps are widely used in computer systems and are commonly used in log recording, data backup, data synchronization, data alignment and other aspects. In an embodiment of the present application, after the vehicle-side incremental parameters and structured data are determined as the driving incremental data, it also includes obtaining the data collection time of the driving incremental data, and timestamping the driving incremental data according to the data collection time to obtain driving incremental data with a timestamp.

[0049] In one embodiment of the present application, a storage time series is generated according to a preset time interval, and driving incremental data with timestamps are aligned within the time series according to the timestamps, so as to store the driving incremental data according to the time series.

[0050] Generate timestamps based on data collection time, annotate data, and align data based on a unified time series for storage, so that data storage can be arranged based on a unified sequence, making it easier to extract and process data at specific time nodes.

[0051] In step S220 , the driving incremental data is classified based on the scenario application type to obtain multiple scenario type data sets.

[0052] In one embodiment of the present application, the above-mentioned scene type data set includes a plurality of driving incremental data.

[0053] In one embodiment of the present application, after the driving incremental data is classified based on the scenario application type, a classification label is generated according to the classification type, and the data in each scenario type data set is marked according to the data classification label.

[0054] In step S230 , metadata is extracted from the plurality of scene type data groups to obtain driving scene metadata.

[0055] In one embodiment of the present application, after metadata extraction is performed on the multiple scene type data groups, the data scene key features of the vehicle-side driving model are obtained, and data retrieval is performed in the extracted initial metadata based on the data scene key features to determine the driving scene metadata.

[0056] In one embodiment of the present application, the above-mentioned data scenario key features include but are not limited to data classification features, data attribute labeling features and scenario features.

[0057] It should be noted that the above-mentioned process of data retrieval of initial metadata based on the key features of the data scenario can be performed by relevant technical personnel, or the model can be trained after the key features of the determined data scenario are annotated based on a deep learning model, and data retrieval and extraction can be performed based on the trained model. The specific data retrieval method is not limited here.

[0058] In one embodiment of the present application, a non-relational database is used to manage driving scene metadata.

[0059] Retrieving data based on the key features of the data scenario can narrow the scope of data requirements based on the training model and reduce the amount of data to be processed, thereby improving data processing efficiency.

[0060] In step S240, the driving scene metadata is converted into a data format according to the pre-configured data processing node, and the training data for the vehicle-side driving model is generated based on the format-converted driving scene metadata.

[0061] In one embodiment of the present application, before converting the data format of the driving scene metadata, the driving scene metadata undergoes data cleaning. The data cleaning includes data deduplication, data fitting, and data desensitization to eliminate invalid data from the driving scene metadata. Deduplication, data fitting, and data desensitization ensure data validity, facilitate further processing of the training data, and ensure the accuracy of the training data.

[0062] In one embodiment of the present application, the topological relationship of the vehicle-side driving module refers to the connection method and communication method between each module. Generally speaking, the vehicle-side driving module includes a perception module, a decision module, and an execution module. In an embodiment of the present application, the data dependency and data index relationship of the driving scene metadata are determined based on the topological association relationship of the vehicle-side preset driving module, and data dependency labels and data index labels are generated according to the data dependency and data index relationship. The driving scene metadata is annotated according to the data dependency labels and data index labels to obtain the annotated driving scene metadata.

[0063] In one embodiment of the present application, based on the topological association relationship of the preset driving module on the vehicle side, after determining the data dependency and data index relationship of the driving scene metadata, data analysis and data association mining can be performed on the driving scene metadata according to the model training requirements to obtain relevant data attribute information and label information of the data association relationship, in addition to the label information after the model training and application scenario changes. It should be noted that the above-mentioned extensible label information can be adjusted and the label information expanded based on the requirements of the vehicle-side model and application scenario training during the actual implementation of this solution. The above-mentioned exemplary examples of extended label types are not all necessary for expansion, but are only examples of optional extended labels.

[0064] In one embodiment of the present application, the labeled driving scene metadata is converted into a data format according to a preset training data format of the driving model to obtain the format-converted driving scene metadata, so as to generate training data for the vehicle-side driving model based on the format-converted driving scene metadata.

[0065] In the above embodiment, data dependencies and index relationships are determined based on the topological relationship of the vehicle-side driving module. This ensures the data's lineage continuity after indexing and dependency labeling, and also facilitates the reflection of topological relationships and the determination of the correlation of data structures when retrieving data.

[0066] In one embodiment of the present application, the format-converted driving scene metadata is divided into a training data set and a verification data set according to a preset division ratio, and the training data set and the verification data set are stored in a preset training data storage area.

[0067] Figure 4 This is a flowchart of a specific method for generating training data as shown in an exemplary embodiment of the present application. This method can be applied to Figure 1 The implementation environment shown is implemented by the vehicle end 101 and the computer device 102 in the implementation environment. It should be understood that the method can also be applied to other exemplary implementation environments and specifically executed by devices in other implementation environments. This embodiment does not limit the implementation environment to which the method is applicable.

[0068] like Figure 4 As shown, in a specific embodiment of the present application, the specific method for generating data to be trained includes three steps: data reception, data screening, and data processing. Among them, data reception specifically includes accessing the data to obtain initial driving incremental data, and performing data analysis on the above-mentioned driving incremental data to obtain driving incremental data for uploading to the cloud.

[0069] In one embodiment of the present application, a meta data package can be obtained after filtering the data. The meta data package is the metadata in the above embodiment. The meta data package is then processed based on nodes 1 to 3 to obtain a meta data package that inherits the basic attributes of the data before processing, as well as labels and attributes extended to the meta data package during the data processing process.

[0070] In a specific embodiment of the present application, the above-mentioned node 1 is to clean the meta data, and the processing result is a subset of the meta data; node 2 is to mine and analyze the meta data, and this process will derive new label information or attribute information to supplement the descriptive information of the meta data; node 3 is to read the file data in the meta data, convert the file, derive new files, and generate data to be trained based on the new files.

[0071] In a specific embodiment of the present application, DataAPI (data resources) provides a database operation interface, and users can interact with the database through HTTPS requests or SDKs, thereby realizing the use of Web service interfaces, reducing the cost of developers managing application servers, and achieving the goal of simplified and efficient development. In this embodiment, dataApi is used for the above three nodes to maintain various types of processing results. For the data cleaning scenario of node 1, the relationship between the node task and the cleaned data index is retained, and only the index information is passed when the data flows backward; for the mining and analysis scenario of node 2, the structured relationship between the node task and the meta data, extended tags or attributes is retained, and the original meta data and extended information are passed to the backward process; for the derivative file scenario of node 3, a new meta data record is created. The file in this data record is a derivative file, and the rest of the information inherits the original meta data.

[0072] Among them, the above nodes are in a series-parallel relationship, and a workflow engine is used to be responsible for resource scheduling, task status management, and process management of the data processing tasks of the above nodes.

[0073] Figure 5 This is a flowchart of a specific driving incremental data acquisition method shown in an exemplary embodiment of the present application. Figure 1 The implementation environment shown is implemented by the vehicle end 101 and the computer device 102 in the implementation environment. It should be understood that the method can also be applied to other exemplary implementation environments and specifically executed by devices in other implementation environments. This embodiment does not limit the implementation environment to which the method is applicable.

[0074] like Figure 5As shown, in a specific embodiment of the present application, step S1 is to store and scan the autonomous driving data after collection, wherein the storage operation adopts the disk insertion operation of connecting the transit server to the data disk and the data center, and adopts the scheduled scanning service to detect the update status of the data disk file.

[0075] Step S2 performs a storage operation on the database. If an update file exists, it is copied to the data center's distributed file system. Simultaneously, the file index information is stored in a non-relational database for file retrieval. The file name of the update file can include information such as the vehicle number and recording time to enrich the file index information. If structured data exists, it is stored in the relational database.

[0076] Step S3 is to parse the updated file. After detecting the file update, a distributed task queue is used to manage the file parsing task, and the message middleware (RabbitMQ) is used to distribute the parsing task of the updated file, and the file is parsed in each task execution unit (Celeryworker). In this specific embodiment, the Celery distributed asynchronous task framework is selected to realize parallel parsing of files to speed up data parsing.

[0077] In step S4, the standard stored file data is parsed to obtain the vehicle information, sensor signals, and other information recorded in the update file. If the update file records camera image streams or lidar data, the image streams are converted to JPG format for storage, and the lidar data is converted to the PCD standard point cloud file format for storage. Data is aligned based on each frame's timestamp, and missing values ​​are fitted and supplemented. The resulting data is aggregated into a single frame of metadata and stored in a non-relational database. This metadata records vehicle information, sensor signals, camera images, point cloud files, and other vehicle-side driving data at a specific point in time.

[0078] A method, device, electronic device and storage medium for generating data to be trained in an embodiment of the present application obtains driving incremental data and scene application types on the vehicle side, uploads the driving incremental data to the cloud, and then classifies the driving incremental data based on the scene application type to obtain multiple scene type data sets, and extracts metadata to obtain driving scene metadata, and then converts the driving scene metadata into a data format according to a pre-configured data processing node, and generates data to be trained for the vehicle-side driving model based on the format-converted driving scene metadata; this solution can effectively avoid the waste of local storage resources on intermediate data by uploading data to the cloud and performing corresponding data operations, and based on the metadata extraction of scene data and the overall labeling and control of data based on metadata, effectively improves the operational efficiency of data integrity control, facilitates the generation of data to be trained required for the vehicle-side model, and also provides beneficial effects including monitoring the initial driving data volume on the vehicle side, determining the data increment, and quantifying the data increment according to the data structure. According to the extraction, the data content in a unified and standardized storage format is obtained, which is convenient for further processing operations on the data and the retrieval and viewing of intermediate data, and can also be stored in a unified manner; a timestamp is generated according to the data collection time, and the data is labeled, and the data is aligned and stored based on a unified time series, so that the data storage can be arranged based on a unified sequence, which is convenient for extracting and processing data at specific time nodes; data dependencies and index relationships are determined based on the topological relationship of the vehicle-side driving module, which can ensure the bloodline continuity of the data after indexing and dependency labeling, and facilitate the reflection of the topological relationship and the determination of the correlation of the data structure when the data is retrieved; after deduplication, data fitting and data desensitization of the metadata, the validity of the data can be guaranteed, which is convenient for further processing of the training data and the accuracy of the data to be trained; data is retrieved according to the key features of the data scenario, which can narrow the data demand range based on the training model and reduce the amount of data to be processed, thereby improving data processing efficiency.

[0079] The following describes an embodiment of the device of the present application, which can be used to execute the method for generating data to be trained in the above-mentioned embodiment of the present application. For details not disclosed in the embodiment of the system of the present application, please refer to the embodiment of the method for generating data to be trained in the above-mentioned embodiment of the present application.

[0080] Figure 6 This is a schematic diagram of a device for generating training data as shown in an exemplary embodiment of the present application. Figure 2 The implementation environment shown in FIG. 1 is specifically configured in the computer device 102. The apparatus may also be applicable to other exemplary implementation environments and specifically configured in other devices. This embodiment does not limit the implementation environment to which the apparatus is applicable.

[0081] like Figure 6 As shown, the exemplary device for generating data to be trained includes: a vehicle-side data acquisition module 601, an incremental data classification module 602, a metadata extraction module 603 and a metadata processing module 604.

[0082] Among them, the vehicle-side data acquisition module 601 is used to obtain the vehicle-side driving incremental data and scene application type; the incremental data classification module 602 is used to classify the driving incremental data based on the scene application type to obtain multiple scene type data sets, and the scene type data sets include multiple driving incremental data; the metadata extraction module 603 is used to extract metadata from the multiple scene type data groups to obtain driving scene metadata; the metadata processing module 604 is used to convert the data format of the driving scene metadata according to the pre-configured data processing node, and generate data to be trained for the vehicle-side driving model based on the driving scene metadata after the format conversion.

[0083] An embodiment of the present application also provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by one or more processors, the electronic device implements the method for generating data to be trained provided in the above-mentioned embodiments.

[0084] Figure 7 This is a schematic diagram of the structure of a computer system of an electronic device according to an exemplary embodiment of the present application. Figure 7 The computer system 700 of the electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0085] like Figure 7 As shown, computer system 700 includes a central processing unit (CPU) 701, which can perform various appropriate actions and processes according to the program stored in read-only memory (ROM) 702 or the program loaded from the storage portion into random access memory (RAM) 703, such as the method described in the above embodiment. Various programs and data required for system operation are also stored in RAM 703. CPU 701, ROM 702 and RAM 703 are connected to each other via a bus. Input / output (I / O) interface 705 is also connected to bus 704.

[0086] The following components are connected to the I / O interface 705: an input section 706 including a keyboard, a mouse, and the like; an output section 707 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage section 708 including a hard disk and the like; and a communication section 709 including a network interface card such as a LAN (Local Area Network) card or a modem. The communication section performs communication processing via a network such as the Internet. A drive is also connected to the I / O interface 705 as needed. Removable media 711, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like, is installed in the drive 710 as needed so that computer programs read therefrom can be installed into the storage section 708 as needed.

[0087] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 709, and / or installed from a removable medium 711. When the computer program is executed by the central processing unit (CPU) 701, the various functions defined in the system of the present application are executed.

[0088] It should be noted that the computer-readable medium shown in the embodiments of the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, for example, an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, device or component, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries a computer-readable computer program. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. A computer program embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, or any suitable combination thereof.

[0089] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. Among them, each box in the flowchart or block diagram can represent a module, program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0090] In the corresponding drawings of the above embodiments, connecting lines can represent the connection relationship between various components to represent more constituent signal paths (constituent_signalpath) and / or one or more ends of some lines have arrows to indicate the main information flow direction. The connecting lines are used as an identifier, not a limitation of the scheme itself, but the use of these lines in combination with one or more exemplary embodiments helps to connect circuits or logic units more easily. Any represented signal (determined by design requirements or preferences) can actually include one or more signals that can be transmitted in any direction and can be implemented with any appropriate type of signal scheme.

[0091] The units involved in the embodiments described in this application may be implemented by software or hardware, and the units described may also be set in a processor. In some cases, the names of these units do not constitute limitations on the units themselves.

[0092] Another aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above. The computer-readable storage medium may be included in the electronic device described in the above embodiments, or may exist independently without being incorporated into the electronic device.

[0093] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiment of the application, the features and functions of two or more modules or units described above can be concretized in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be concretized.

[0094] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiments of the present application.

[0095] It should be noted that the present application can be used in a wide variety of general-purpose or special-purpose computing system environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above, and the like.

[0096] Those skilled in the art will readily conceive of other embodiments of the present application after considering the specification and practicing the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of this application and include common knowledge or customary techniques in the art that are not disclosed herein.

[0097] It should be understood that the above content is only a preferred exemplary embodiment of the present application and is not intended to limit the implementation scheme of the present application. Ordinary technicians in this field can easily make corresponding changes or modifications based on the main concept and spirit of the present application. Therefore, the scope of protection of the present application should be the scope of protection required by the claims.

Claims

1. A method for generating data to be trained, characterized in that: The method for generating data to be trained includes: Obtaining incremental driving data and scenario application types from the vehicle side, and uploading the incremental driving data to the cloud; Classifying the driving incremental data based on the scenario application type to obtain a plurality of scenario type data sets, wherein the scenario type data sets include a plurality of driving incremental data; Extracting metadata from the multiple scene type data groups to obtain initial metadata, acquiring key features of the data scene of the vehicle-side driving model, and performing data retrieval in the extracted initial metadata based on the key features of the data scene to obtain driving scene metadata; Performing data format conversion on the driving scene metadata according to a preconfigured data processing node, including determining data dependency and data index relationships of the driving scene metadata based on a topological association relationship of a preset driving module on the vehicle side; generating data dependency labels and data index labels according to the data dependency and data index relationships, and annotating the driving scene metadata according to the data dependency labels and data index labels to obtain annotated driving scene metadata; performing data format conversion on the annotated driving scene metadata according to a preset training data format of a driving model to obtain format-converted driving scene metadata, and generating to-be-trained data for the vehicle-side driving model based on the format-converted driving scene metadata; Based on the driving scene metadata after the format conversion, data to be trained for the vehicle-side driving model is generated.

2. The method for generating training data according to claim 1, wherein: Acquiring incremental driving data on the vehicle side includes: Acquire the initial driving data volume of the vehicle side, and monitor the total amount of data of the initial driving data volume of the vehicle side; If the total amount of data increases, extracting initial incremental data and determining a structure type of the initial incremental data, the structure type including unstructured data and structured parameters; If the incremental data is unstructured data, extracting driving parameters from the unstructured data according to a preset data storage format to obtain vehicle-side incremental parameters; The vehicle-side incremental parameters and structured data are determined as the driving incremental data and stored in a preset incremental data storage area.

3. The method for generating training data according to claim 2, wherein: After determining the vehicle-side incremental parameters and structured data as the driving incremental data, the method for generating data to be trained further includes: Acquiring a data collection time of the driving incremental data, and timestamping the driving incremental data according to the data collection time to obtain the driving incremental data with the timestamp; A storage time series is generated according to a preset time interval, and data alignment is performed on driving incremental data with time stamps within the time series according to the time stamps, so as to store the driving incremental data according to the time series.

4. The method for generating training data according to claim 1, wherein: Before determining the data dependency and data index relationship of the driving scene metadata based on the topological association relationship of the vehicle-side preset driving modules, the method for generating the training data further includes: The driving scene metadata is subjected to data cleaning processing, wherein the data cleaning processing includes data deduplication, data fitting, and data desensitization, so as to eliminate invalid data from the driving scene metadata.

5. The method for generating training data according to claim 1, wherein: Generating the training data required for the vehicle-side driving model based on the driving scene metadata converted from the format includes: The format-converted driving scene metadata is divided into a training data set and a verification data set according to a preset division ratio, and the training data set and the verification data set are stored in a preset training data storage area.

6. A device for generating data to be trained, characterized in that: The device for generating data to be trained comprises: The vehicle-side data acquisition module is used to obtain the vehicle-side driving incremental data and scenario application types; an incremental data classification module, configured to classify the driving incremental data based on the scenario application type to obtain a plurality of scenario type data sets, wherein the scenario type data sets include a plurality of driving incremental data; a metadata extraction module, configured to extract metadata from the plurality of scene type data groups to obtain initial metadata, acquire key features of the data scene of the vehicle-side driving model, and perform data retrieval in the extracted initial metadata based on the key features of the data scene to obtain driving scene metadata; A metadata processing module is used to perform data format conversion on the driving scene metadata according to a pre-configured data processing node, including determining the data dependency and data index relationship of the driving scene metadata based on the topological association relationship of the vehicle-side preset driving module; generating data dependency labels and data index labels according to the data dependency and data index relationship, and annotating the driving scene metadata according to the data dependency labels and data index labels to obtain annotated driving scene metadata; performing data format conversion on the annotated driving scene metadata according to a preset training data format of the driving model to obtain format-converted driving scene metadata, and generating to-be-trained data for the vehicle-side driving model based on the format-converted driving scene metadata; and generating to-be-trained data for the vehicle-side driving model based on the format-converted driving scene metadata.

7. An electronic device, characterized in that: include: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the electronic device to implement the method for generating data to be trained as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that Computer-readable instructions are stored thereon, and when the computer-readable instructions are executed by a processor of a computer, the computer is caused to execute the method for generating data to be trained according to any one of claims 1 to 5.

Images