A vehicle use scenario expansion method, device, equipment and vehicle

Abstract

The application relates to the technical field of intelligent vehicles, and discloses a vehicle use scene expansion method, device, equipment and vehicle. The vehicle use scene expansion method comprises the following steps: a scene feature expansion network is constructed, and the scene feature expansion network is trained according to artificially marked real scene features; and the vehicle use scene data is expanded based on the trained scene feature expansion network, so that the expansion of the vehicle use scene data can be completed through a small amount of marked real scene data, and the diversity and richness of the vehicle use scene data are improved.

Description

Technical Field

[0001] This application relates to the field of intelligent vehicle technology, and in particular to a method, device, equipment and vehicle for expanding vehicle usage scenarios. Background Technology

[0002] Multi-domain fusion scenario recommendation technology is a comprehensive technology that combines user behavior and preferences with information from multiple domains within a specific scenario to provide personalized recommendations. It aims to improve the accuracy and diversity of recommendation systems by continuously collecting and processing data, training and optimizing models, and providing feedback and evaluation to deliver a more precise, richer, and more personalized user experience.

[0003] Multi-domain fusion scenario recommendation technology based on vehicle usage scenarios currently relies heavily on manual data collection. This involves manually evaluating and tagging vehicle status, voice, video, and image data collected from the vehicle, creating a unified set of scenario data. Manual data collection suffers from high costs, insufficient data richness, and imbalanced scenario data. Summary of the Invention

[0004] This application provides a method, apparatus, equipment, and vehicle for expanding vehicle usage scenarios, in order to solve one or more technical problems existing in the prior art, and at least provide a beneficial option or create conditions.

[0005] One aspect of this application provides a method for expanding vehicle usage scenarios, comprising the following steps: obtaining real scene features based on vehicle usage scenarios and manually labeling the real scene features, wherein the real scene features are generated through multi-dimensional feature fusion; constructing a scene feature expansion network, the scene feature expansion network including a feature generator and a discriminator; training the scene feature expansion network according to the manually labeled real scene features; and expanding vehicle usage scenario data based on the trained scene feature expansion network.

[0006] Some technical solutions of this application, wherein obtaining real scene features based on vehicle usage scenarios includes: acquiring the current vehicle usage scenario; extracting vehicle status features, personnel intention features, and audio-visual features according to the vehicle usage scenario; concatenating the vehicle status features, personnel intention features, and audio-visual features; and generating real scene features based on the concatenated features.

[0007] Some technical solutions of this application, wherein training the scene feature extension network based on the manually labeled real scene features includes: the feature generator obtaining generated scene features based on latent spatial parameters and random noise; and training the scene feature extension network based on the generated scene features and the real scene features.

[0008] In some technical solutions of this application, the feature generator obtains generated scene features based on latent space parameters and random noise, including: inputting initial latent variables into a mapping network to obtain the latent space parameters, and fusing the latent space parameters together with random noise in an adaptive instance normalization manner into the feature generator to obtain generated scene features; wherein, the random noise is used to control the variation of the fineness of the generated scene features.

[0009] Some technical solutions of this application, wherein training the scene feature extension network based on the generated scene features and the real scene features includes: inputting the generated scene features and the real scene features into the discriminator; the discriminator performing category judgment and true / false judgment on the generated scene features based on the real scene feature information; continuously training the feature generator and the discriminator alternately; when the discriminator learns sufficient ability to distinguish between true / false and category, and the feature generator learns sufficient ability to generate scene features that can deceive the discriminator, the training of the scene feature extension network is terminated.

[0010] Some technical solutions of this application, wherein the scene feature augmentation network trained on the scene is used to augment the vehicle use scene data, include: fixing the parameters of the feature generator and the discriminator; adjusting the trainable latent variables according to the real scene features; adjusting the random noise at different levels in the feature generator, and obtaining multiple generated scene features based on the adjusted latent variables; wherein the multiple generated scene features are consistent with the category labels of the real scene features.

[0011] According to some technical solutions of this application, the construction of the scene feature augmentation network includes: constructing the scene feature augmentation network based on StyleGAN, a generative adversarial network for style transformation.

[0012] Another aspect of this application provides a vehicle usage scenario expansion device, which includes: a scenario feature generation module, used to obtain real scenario features based on a vehicle usage scenario and manually label the real scenario features, wherein the real scenario features are generated through multi-dimensional feature fusion; an expansion network construction module, used to construct a scenario feature expansion network, the scenario feature expansion network including a feature generator and a discriminator; a training module, used to train the scenario feature expansion network according to the manually labeled real scenario features; and a scenario data expansion module, used to expand the vehicle usage scenario data based on the trained scenario feature expansion network.

[0013] Another aspect of this application provides an in-vehicle electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for expanding vehicle usage scenarios.

[0014] Another aspect of this application provides a vehicle that includes on-board electronic equipment as described in the other aspect above.

[0015] This application has at least the following beneficial effects:

[0016] The vehicle usage scenario expansion method of this application constructs a scenario feature expansion network and trains the scenario feature expansion network based on manually labeled real scenario features. Based on the trained scenario feature expansion network, the vehicle usage scenario data can be expanded with a small amount of labeled real scenario data, thereby improving the diversity and richness of the vehicle usage scenario data.

[0017] It is understood that the beneficial effects and methods of the devices, equipment and vehicle technical solutions disclosed in this application are the same, and will not be repeated here.

[0018] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0019] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0020] Figure 1 This is a flowchart illustrating the method for expanding vehicle usage scenarios according to an embodiment of this application.

[0021] Figure 2 A flowchart illustrating the process of extending the network with scene features according to an embodiment of this application;

[0022] Figure 3 This is a flowchart illustrating the method for obtaining real-world scene features according to an embodiment of this application.

[0023] Figure 4 This is a schematic block diagram of the vehicle usage scenario expansion device according to an embodiment of this application;

[0024] Figure 5 This is a schematic diagram of the hardware structure of an in-vehicle electronic device according to an embodiment of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0026] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various concepts, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to a determination” as used herein may be interpreted as “when…” or “when…” or “in response to a determination.”

[0027] As used in this application, the terms "at least one", "multiple", "each", "any", etc., "at least one" includes one, two or more, "multiple" includes two or more, "each" refers to each of the corresponding multiples, and "any" refers to any one of the multiples.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0029] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0030] To make the inventive concept of the technical solution of this application easier to understand, the related technologies involved in the technical solution of this application will now be introduced.

[0031] Multi-domain fusion scenario recommendation technology is a comprehensive technology that combines user behavior and preferences with multi-domain information in a specific scenario for personalized recommendations. It aims to improve the accuracy and diversity of recommendation systems by continuously collecting and processing data, training and optimizing models, and providing feedback and evaluation to deliver a more accurate, richer, and more personalized user experience. A typical process for multi-domain fusion scenario recommendation technology to execute intelligent recommendations on in-vehicle systems is as follows:

[0032] Data collection and processing: By collecting user behavior data and content data (such as voice, images, videos, etc.), we perform preprocessing, cleaning, and labeling to provide a high-quality dataset for subsequent model training.

[0033] User profiling and preference analysis: By analyzing user behavior data, we extract user interest characteristics, construct user profiles, and further analyze user preferences. This helps us understand user preferences, needs, and habits, providing important basis for subsequent recommendations.

[0034] Domain knowledge fusion: This involves integrating knowledge from different domains, such as user information and contextual information, to construct a cross-domain semantic network. Through cross-domain knowledge fusion, richer contextual information can be obtained, improving the accuracy of recommendations.

[0035] Feature extraction and scene-aware model training: Using deep learning methods, user and scene features are extracted from the data, and these features are used to train a scene-aware model.

[0036] Scene-aware recommendation: In specific real-world scenarios, based on user behavior and preferences, and combined with multi-domain information, a trained model determines the scene category and delivers scene-specific recommended content to the in-vehicle infotainment system. Specifically, the scene feature determination process can be described as follows: the scene feature dataset is input into a feature classification network, which then provides a classification result based on the different scene features. For example, scene classification results could include: children getting into the car, pets getting into the car, refueling, and traffic jams.

[0037] Feedback and Optimization: During the recommendation process, user feedback information, such as reach rate and browsing time, is collected to evaluate the accuracy and effectiveness of the recommendations. Then, the recommendation model is adjusted and optimized based on the feedback information to improve the accuracy of the recommendations.

[0038] Data collection and processing form the foundation of the entire technical solution. The quality of the collected data directly affects the training effectiveness of feature extraction and scene perception models, influencing the final model's scene judgment and recommendation results.

[0039] Currently, data collection typically involves manual methods to assess and label the collected vehicle status and information from various domains, such as voice, video, and images, with corresponding scene tags. This information from multiple domains within the same scene collectively constitutes a unified set of scene data, which is then used for model training. However, manually collecting scene data currently has the following drawbacks:

[0040] 1. It involves many data fields, and the cost of manually collecting and labeling data is too high;

[0041] 2. Insufficient data richness in special scenarios. For example, intelligent recommendations in pet mode require image content related to different categories of pets;

[0042] 3. Imbalanced dataset. Because the amount of data required for model training is very large, with the current manual collection capabilities, it is difficult to collect enough training data in some scenarios, which will affect the final recognition performance of the model.

[0043] In view of this, this application provides a method, apparatus, equipment, and vehicle for expanding vehicle usage scenarios to solve the above problems.

[0044] The following is a description of embodiments of this application.

[0045] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating the vehicle usage scenario expansion method according to an embodiment of this application, including the following steps: obtaining real scene features based on the vehicle usage scenario; manually labeling the real scene features, where the real scene features are generated through multi-dimensional feature fusion; constructing a scene feature expansion network, which includes a feature generator and a discriminator; training the constructed scene feature expansion network based on the manually labeled real scene features; and then expanding the vehicle usage scenario data based on the trained scene feature expansion network.

[0046] Specifically, before expanding the data on driving scenarios, a reliable dataset needs to be created first. This dataset should include manually labeled real-world scenario features. To achieve more accurate driving scenario recommendations, these real-world scenario features include multi-dimensional features. For example, real-world scenario features may include audio / video features, user operation intent features, and vehicle status features. By fusing these multi-dimensional features, or multi-domain feature fusion, the driving scenarios can be more richly and meticulously depicted, leading to more accurate driving scenario recommendations. For example, driving scenarios may include scenarios such as children getting into the car, pets getting into the car, refueling, highway driving, and traffic jams.

[0047] In some embodiments of the present invention, the scene data of vehicle use is expanded by constructing a scene feature expansion network. Specifically, the scene feature expansion network can be trained using manually labeled real scene features. The scene feature expansion network includes a feature generator and a discriminator. The feature generator generates scene features, and the discriminator determines the scene features generated by the feature generator and their categories based on the real scene features. By training the feature generator and discriminator of the scene feature expansion network, a scene feature expansion network capable of effectively expanding the scene data of vehicle use is obtained. Then, the scene data of vehicle use is expanded based on the trained scene feature expansion network.

[0048] In some embodiments of the present invention, the vehicle usage scenario expansion method further includes obtaining real scene features based on the vehicle usage scenario. The steps for obtaining real scene features include: first, identifying the current vehicle usage scenario; and then further extracting vehicle status features, occupant intent features, and audio / video features based on the current vehicle usage scenario. Subsequently, the vehicle status features, occupant intent features, and audio / video features are subjected to feature concatenation, and real scene features are generated based on the concatenated features.

[0049] For details, please refer to Figure 3 , Figure 3This is a flowchart illustrating the method for obtaining real-world scene features according to an embodiment of this application. It is for illustrative purposes only. When the current driving scenario is identified as a safe exit door opening warning scenario, it is necessary to extract vehicle status features, occupant intention features, and visual features within the vehicle. Vehicle status features may include the current gear position and speed; occupant intention features may include seat adjustment, door status, and seatbelt engagement / disengagement status; visual features within the vehicle may include the results of target detection in the image behind the vehicle. It should be understood that the extraction targets for vehicle status features, occupant intention features, and audio-visual features within the vehicle will differ depending on the identified driving scenario. For example, in the safe exit door opening warning scenario, visual feature extraction includes the results of target detection in the image behind the vehicle; in this case, audio features may not be extracted. However, if switching to a traffic jam scenario, audio features will need to be extracted, and the extracted visual features will also differ. Specifically, which features should be extracted in which scenarios can be flexibly set based on the actual scenario, historical experience data, and user feedback. After extracting the vehicle status features, personnel intention features, and visual features on the vehicle for the corresponding scene, scene feature fusion is performed on these features. The fusion process involves dimensional splicing of the features to generate real scene features. Then, the real scene features are manually labeled, also known as scene category labeling, thus forming labeled training data.

[0050] In some embodiments of the present invention, the scene feature extension network is trained based on artificially labeled real scene features. This includes a feature generator obtaining generated scene features based on latent spatial parameters and random noise, and training the scene feature extension network based on the generated scene features and real scene features.

[0051] For details, please refer to Figure 2 , Figure 2 This is a flowchart illustrating the scene feature augmentation network of an embodiment of this application. The scene feature augmentation network includes a feature generator and a discriminator, wherein the feature generator, discriminator, and mapping network are all represented by dashed boxes in conjunction with corresponding diagrams. Specifically, the latent variable Z first generates latent spatial parameters W through the mapping network, and W is input into the feature generator and combined with random noise B1, B2, and B3. Figure 2 The example only shows 3 groups of noise, but in practice, multiple groups can be set. The specific number of noise groups can be set according to the specific structure of the feature generator to generate scene features. The generated scene features and the real scene features are simultaneously input into the discriminator, which performs real-world discrimination and category discrimination on the generated scene features.

[0052] As an example, the feature generator takes an initial latent variable Z as input and generates latent space parameters W after affine transformation through a mapping network. W is then fused into the feature generator along with random noise Bn, used to control style variations, using adaptive instance normalization, to obtain the generated scene features. The initial latent variable Z, also called the latent factor Z, is typically sampled from a normal or uniform distribution and determines the type and style of the generated scene features. The mapping network then transforms the latent factor Z into intermediate latent space parameters W. The mapping network aims to create independent features so that the feature generator can more easily perform rendering while avoiding feature combinations not present in the training dataset. For example, the mapping network uses eight fully connected layers to transform Z into latent space parameters W. By incorporating the latent space parameters W and random noise Bn into the scene feature generation process, and by adjusting Bn at different levels, scene features of varying granularity can be generated without changing the generated scene category. This significantly expands the training dataset required for scene feature judgment and improves the capabilities of the feature classification network.

[0053] The discriminator takes fake scene features generated by the feature generator and manually labeled real scene features as input. Through multi-head output, it uses real scene feature information to simultaneously judge the generated scene features as true / false and classify them. The feature generator and discriminator compete with each other during training to improve their respective capabilities.

[0054] The training of the scene feature augmentation network consists of two stages. In the first stage, the discriminator is fixed, and the feature generator is trained. During this stage, the feature generator continuously attempts to generate feature information that more closely resembles the real scene in order to deceive the discriminator. The second stage involves fixing the generator and training the discriminator. In this stage, the discriminator strives to improve its ability to distinguish between real scene feature information and generated scene feature information. Through this training process, the generator eventually becomes able to generate generated scene features that are almost indistinguishable from real scene feature information. In other words, as the training process improves, the discriminator learns sufficient ability to distinguish between true and false information and different categories, while the generator learns sufficient ability to generate scene features that have fooled the discriminator. At this point, the training process of the scene feature augmentation network is complete.

[0055] In some embodiments of the present invention, the augmentation of vehicle usage scenario data based on the trained scene feature augmentation network mainly includes the following operations: First, the parameters of the feature generator and discriminator are fixed, and the trainable latent variable Z is adjusted according to the real scene features. The random noise Bn at different levels in the feature generator is also adjusted, and multiple generated scene features are obtained based on the adjusted latent variables. It should be understood that the category labels of the multiple generated scene features and the real scene features are consistent.

[0056] Specifically, to generate labeled scene features using a feature generator, the parameters of the generator and discriminator must first be fixed, and the desired augmentation data of the real scene must be input. Then, the scene feature augmentation network adjusts the trainable latent variable Z according to the generation target. The adjusted latent variable is saved and denoted as Z'. Z' is used as the latent variable corresponding to this type of real scene feature. By adjusting the specific values ​​of random noise at different levels in the feature generator, different generated scene features are obtained. It should be understood that the feature generator consists of multiple layers, each a convolutional layer. Through this operation, scene features can be augmented while preserving the original real feature information. The class labels of the newly obtained generated scene features are consistent with the class labels of the real scene features.

[0057] In some embodiments of the present invention, a scene feature augmentation network can be constructed based on StyleGAN, a generative adversarial network that performs style transformations. By using StyleGAN as a scene feature generator, in scenarios with insufficient data richness, scene features similar to the original data features can be effectively obtained by making subtle adjustments to the latent space parameters W of StyleGAN, thereby expanding the training data of the scene perception model.

[0058] The vehicle usage scenario expansion method provided in this application embodiment can expand the dataset in the scenario feature judgment process from the previous dataset to a dataset of real scenario features plus generated scenario features, that is, the generated scenario features supplement the original small amount of manually labeled real scenario features.

[0059] In addition to the vehicle usage scenario expansion method provided above, this application embodiment also provides a vehicle usage scenario expansion device, such as... Figure 4 As shown, Figure 4 This is a schematic block diagram of the vehicle usage scenario expansion device according to an embodiment of this application. The vehicle usage scenario expansion device includes:

[0060] The scene feature generation module is used to obtain real scene features based on the vehicle usage scenario and manually label the real scene features. Here, the real scene features are generated through multi-dimensional feature fusion.

[0061] The extended network construction module is used to build a scene feature extension network, which includes a feature generator and a discriminator.

[0062] The training module is used to train the constructed scene feature extension network based on manually labeled real scene features.

[0063] The scene data augmentation module is used to augment the vehicle usage scene data based on the trained scene feature augmentation network.

[0064] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0065] This application also provides an in-vehicle electronic device, such as... Figure 5 As shown, the in-vehicle electronic equipment includes a memory, one or more processors ( Figure 5 Only one is shown in the diagram, along with a computer program stored in memory and executable on the processor, and a communication interface. The memory stores software programs and units, and the processor executes these stored software programs and units to perform various functional applications and data processing to obtain resources corresponding to the aforementioned preset events. Specifically, the processor implements the aforementioned vehicle usage scenario expansion method by running the aforementioned computer program stored in memory. This in-vehicle electronic device can include any intelligent terminal such as an in-vehicle computer.

[0066] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0067] This application also provides a vehicle, which includes the in-vehicle electronic equipment described in the above embodiments. It should be understood that the vehicle may include autonomous vehicles, driverless vehicles, and manned vehicles, and is not limited to any specific model. Any vehicle that can directly or indirectly implement the methods of the embodiments in this application should be included within the scope of vehicles covered by this application.

[0068] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for expanding vehicle usage scenarios.

[0069] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0070] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0071] This application also provides a computer program product, which includes a computer program that, when executed by one or more processors, can implement the steps of the vehicle usage scenario expansion method described above.

[0072] The vehicle usage scenario expansion method, device, in-vehicle electronic device, vehicle, storage medium, and computer program product provided in this application embodiment construct a scenario feature expansion network and train the scenario feature expansion network based on manually labeled real scenario features. Based on the trained scenario feature expansion network, vehicle usage scenario data can be expanded using a small amount of labeled real scenario data, significantly reducing labor costs. Utilizing StyleGAN as a scenario feature generator, in scenarios with insufficient data richness, subtle adjustments to the latent space parameter W of StyleGAN can effectively obtain scenario features similar to the original data features, expanding the training data of the scenario perception model and generating a smaller number of scenario category data features. This effectively solves the data imbalance problem and mitigates the negative factors caused by data imbalance in the training of the scenario perception model.

[0073] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0074] Although specific embodiments have been described herein, those skilled in the art will recognize that many other modifications or alternative embodiments are also within the scope of this disclosure. For example, any of the functions and / or processing capabilities described in connection with a particular device or component can be performed by any other device or component. Furthermore, while various exemplary embodiments and architectures have been described according to embodiments of this disclosure, those skilled in the art will recognize that many other modifications to the exemplary embodiments and architectures herein are also within the scope of this disclosure.

[0075] The foregoing description of certain aspects of this disclosure refers to block diagrams and flowcharts of systems, methods, systems, and / or computer program products according to exemplary embodiments. It should be understood that one or more blocks in the block diagrams and flowcharts, as well as combinations of blocks in the block diagrams and flowcharts, can be implemented by executing computer-executable program instructions, respectively. Similarly, according to some embodiments, some blocks in the block diagrams and flowcharts may not need to be executed in the order shown, or may not all need to be executed. Furthermore, additional components and / or operations beyond those shown in the blocks in the block diagrams and flowcharts may exist in some embodiments.

[0076] Therefore, blocks in block diagrams and flowcharts support combinations of means for performing a specified function, combinations of elements or steps for performing a specified function, and program instruction means for performing a specified function. It should also be understood that each block in a block diagram and flowchart, and combinations of blocks in block diagrams and flowcharts, can be implemented by a dedicated hardware computer system or a combination of dedicated hardware and computer instructions that performs a specific function, element, or step.

[0077] The program modules, applications, etc., described herein may include one or more software components, including, for example, software objects, methods, data structures, etc. Each such software component may include computer-executable instructions that, in response to execution, cause at least a portion of the functionality of this document (e.g., one or more operations of the exemplary methods described herein) to be performed.

[0078] Software components can be coded using any of a variety of programming languages. An exemplary programming language could be a low-level programming language, such as assembly language associated with a specific hardware architecture and / or operating system platform. Software components including assembly language instructions may need to be converted into executable machine code by an assembler before being executed by the hardware architecture and / or platform. Another exemplary programming language could be a higher-level programming language that is portable across multiple architectures. Software components including higher-level programming languages ​​may need to be converted into an intermediate representation by an interpreter or compiler before execution. Other examples of programming languages ​​include, but are not limited to, macro languages, shell or command languages, job control languages, scripting languages, database query or search languages, or report writing languages. In one or more exemplary embodiments, a software component containing instructions from one of the above-described programming language examples can be executed directly by the operating system or other software components without first being converted into another form.

[0079] Software components can be stored as files or other data storage structures. Software components of similar type or related function can be stored together in a specific directory, folder, or library. Software components can be static (e.g., pre-defined or fixed) or dynamic (e.g., created or modified at runtime).

[0080] The embodiments of this application have been described in detail above with reference to the accompanying drawings. However, this application is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of this application.

Claims

1. A method for expanding vehicle usage scenarios, characterized in that, Includes the following steps: Based on the vehicle usage scenario, real scene features are obtained and manually labeled. The real scene features are generated through multi-dimensional feature fusion. A scene feature augmentation network is constructed, which includes a feature generator and a discriminator; The initial latent variables are input into the mapping network to obtain the latent space parameters; The latent space parameters are fused with random noise in an adaptive instance normalization manner into the feature generator to obtain generated scene features, wherein the random noise is used to control the variation of the fineness of the generated scene features; The generated scene features and the real scene features are input into the discriminator; The discriminator performs category and authenticity judgments on the generated scene features based on real scene feature information; The feature generator and the discriminator are trained alternately and continuously. When the discriminator learns the ability to distinguish between true and false and between different categories, and the feature generator learns the ability to generate scene features that can fool the discriminator, the training of the scene feature augmentation network ends. The scene feature augmentation network, after training, is used to augment the vehicle usage scene data.

2. The method for expanding vehicle usage scenarios according to claim 1, characterized in that, The real-world scene features obtained based on vehicle usage scenarios include: Obtain the current vehicle usage scenario; Based on the vehicle usage scenario, extract the vehicle's status features, the person's intention features, and the audio and video features respectively; The vehicle's status features, the people's intention features, and the audio and video features are spliced ​​together. Real-world scene features are generated based on the spliced ​​features.

3. The method for expanding vehicle usage scenarios according to claim 1, characterized in that, The scene feature augmentation network, trained after training, augments the vehicle usage scenario data, including: The parameters of the feature generator and the discriminator are fixed; Adjust the trainable latent variables based on the real-world scene features; Adjust the random noise at different levels in the feature generator, and obtain multiple generated scene features based on the adjusted latent variables; Among them, multiple generated scene features have the same category labels as the real scene features.

4. The method for expanding vehicle usage scenarios according to claim 1, characterized in that, The construction of the scene feature augmentation network includes: The scene feature augmentation network is constructed based on StyleGAN, a generative adversarial network that transforms style.

5. A vehicle usage scenario expansion device, characterized in that, The vehicle usage scenario expansion device includes: The scene feature generation module is used to obtain real scene features based on the vehicle usage scenario and manually label the real scene features, wherein the real scene features are generated through multi-dimensional feature fusion. An extended network construction module is used to construct a scene feature extended network, wherein the scene feature extended network includes a feature generator and a discriminator; The training module is used to input initial latent variables into the mapping network to obtain latent space parameters; to fuse the latent space parameters with random noise in an adaptive instance normalization manner into the feature generator to obtain generated scene features, wherein the random noise is used to control the variation of the fineness of the generated scene features; to input the generated scene features and the real scene features into the discriminator; the discriminator performs category judgment and true / false judgment on the generated scene features based on the real scene feature information; to continuously train the feature generator and the discriminator alternately; when the discriminator learns the ability to distinguish between true / false and category, and the feature generator learns the ability to generate scene features that can deceive the discriminator, the training of the scene feature augmentation network ends; The scene data augmentation module is used to augment the vehicle usage scene data based on the trained scene feature augmentation network.

6. A vehicle-mounted electronic device, characterized in that, The vehicle-mounted electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the vehicle usage scenario expansion method as described in any one of claims 1 to 4.

7. A vehicle, characterized in that, The vehicle includes the on-board electronic equipment as described in claim 6.

Images