Method, device, equipment and readable storage medium for obtaining map construction data

Abstract

The present application discloses a method, device, equipment and readable storage medium for obtaining map construction data, and belongs to the field of image processing technology. The method includes: obtaining a second label that meets the task requirements based on a first label of the first image data; generating second image data that matches the first image data based on the second label; obtaining a first neural network model based on training the second image data; performing style transfer on the third image data through the first neural network model to obtain map construction data. Even in the absence of the first image data, the present application can still use the generated second image data for model training, thereby improving the performance of the trained neural network model. Therefore, when the third image data is style transferred based on the first neural network model, the image quality of the obtained map construction data is higher and the accuracy is improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of image processing technology, and in particular to a method, device, equipment and readable storage medium for obtaining map construction data. Background Art

[0002] With the development of social technology, people use maps more and more frequently in their daily lives, and the demand for maps is increasing. In order to meet people's demand for maps in various life scenarios, it is necessary to use map automation production technology to complete map construction in different life scenarios. To achieve map automation production, it is necessary to obtain map construction data, such as street view image data.

[0003] In the related art, map construction data is obtained by annotating the map construction data and synthesizing the image style. Among them, annotation is mainly achieved through manual annotation; synthesis is mainly achieved by generating an adversarial network or encoder through image data in open source datasets, and using the adversarial network or encoder to transfer the image style of the original image data to the scene where the map construction data is located to obtain the map construction data with the image style of the scene.

[0004] Since the amount of map construction data required to build a map road network is large, manual annotation is costly and inefficient; if the image data in the open source dataset is missing, the performance of the generated adversarial network or encoder is poor, the image quality generated by the adversarial network or encoder is poor, and the accuracy of the map construction data finally obtained is low. Summary of the invention

[0005] The embodiments of the present application provide a method, device, equipment and readable storage medium for obtaining map construction data, which can be used to solve the problems in related technologies. The technical solution is as follows:

[0006] On the one hand, an embodiment of the present application provides a method for obtaining map construction data, the method comprising:

[0007] Acquire a second label that meets the task requirements based on the first label of the first image data;

[0008] generating second image data matching the first image data based on the second tag;

[0009] A first neural network model is obtained by training based on the second image data;

[0010] The style of the third image data is transferred through the first neural network model to obtain map construction data.

[0011] In a possible implementation manner, the step of obtaining a first neural network model based on the second image data training includes:

[0012] Determine the style transfer type;

[0013] A first neural network model is obtained by training based on the second image data and the information of the style transfer type.

[0014] In a possible implementation, performing style migration on the third image data by using the first neural network model to obtain map construction data includes:

[0015] The third image data and information about the style transfer type are input into the first neural network model, and the style of the third image data is transferred through the first neural network model to obtain map construction data with the style transfer type.

[0016] In a possible implementation, acquiring a second label that meets task requirements based on the first label of the first image data includes:

[0017] A second neural network model is obtained by training based on the first noise and the first label;

[0018] A second noise is input into the second neural network model to obtain a second label that meets the task requirements.

[0019] In a possible implementation manner, the generating, based on the second tag, second image data matching the first image data includes:

[0020] acquiring the first image data;

[0021] A third neural network model is obtained by training based on the first label and the first image data;

[0022] The second label is input into the third neural network model, and the second image data is output through the third neural network model.

[0023] In a possible implementation manner, the step of obtaining a first neural network model based on the second image data training includes:

[0024] A first neural network model is obtained by training based on the first image data and the second image data.

[0025] In a possible implementation manner, the step of obtaining a first neural network model based on the second image data training includes:

[0026] Determine the style transfer type;

[0027] A first neural network model is obtained by training based on the first image data, the second image data and information of the style transfer type.

[0028] In another aspect, a device for acquiring map construction data is provided, the device comprising:

[0029] An acquisition module, used for acquiring a second label that meets the task requirements based on the first label of the first image data;

[0030] a data generating module, configured to generate second image data matching the first image data based on the second tag;

[0031] A training module, used for training a first neural network model based on the second image data;

[0032] The style transfer module is used to perform style transfer on the third image data through the first neural network model to obtain map construction data.

[0033] In a possible implementation, the training module is used to determine the style transfer type; and to obtain a first neural network model through training based on the second image data and information about the style transfer type.

[0034] In one possible implementation, the style transfer module is used to input the third image data and the information of the style transfer type into the first neural network model, perform style transfer on the third image data through the first neural network model, and obtain map construction data with the style transfer type.

[0035] In one possible implementation, the acquisition module is used to acquire a first label of the first image data; train a second neural network model based on the first noise and the first label; and input a second noise into the second neural network model to obtain a second label that meets task requirements.

[0036] In one possible implementation, the data generation module is used to obtain the first image data; train a third neural network model based on the first label and the first image data; input the second label to the third neural network model, and output the second image data through the third neural network model.

[0037] In a possible implementation, the training module is used to train a first neural network model based on the first image data and the second image data.

[0038] In a possible implementation, the training module is used to determine a style transfer type; and to obtain a first neural network model through training based on the first image data, the second image data, and information about the style transfer type.

[0039] On the other hand, a computer device is provided, comprising a processor and a memory, wherein at least one computer program is stored in the memory, and the at least one computer program is loaded and executed by the processor so that the computer device implements any of the above-mentioned methods for obtaining map construction data.

[0040] On the other hand, a computer-readable storage medium is also provided, in which at least one computer program is stored. The at least one computer program is loaded and executed by a processor so that a computer implements any of the above-mentioned methods for obtaining map construction data.

[0041] On the other hand, a computer program product or a computer program is also provided, the computer program product or the computer program comprising computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs any of the above methods for obtaining map construction data.

[0042] The technical solution provided by the embodiments of the present application brings at least the following beneficial effects:

[0043] By obtaining a second label that meets the task requirements, second image data matching the first image data is generated based on the second label. Since the second label is obtained based on the first label of the first image data, and the second image data matches the first image data, the second image data can be made similar to the first image data. On this basis, the second image data is used to train the first neural network model. Even if the first image data is missing, the generated second image data can still be used for model training, which improves the performance of the trained neural network model. Therefore, when the style of the third image data is transferred based on the first neural network model, the image quality of the obtained map construction data is higher and the accuracy is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0045] Figure 1 It is a schematic diagram of an implementation environment provided by an embodiment of the present application;

[0046] Figure 2 This is a flow chart of a method for obtaining map construction data provided by an embodiment of the present application;

[0047] Figure 3 It is a quality difference map of map construction data provided by an embodiment of the present application;

[0048] Figure 4 This is a flowchart of obtaining map construction data provided by an embodiment of the present application;

[0049] Figure 5 It is a structural schematic diagram of a device for obtaining map construction data provided by an embodiment of the present application;

[0050] Figure 6 It is a structural diagram of a server provided in an embodiment of the present application;

[0051] Figure 7 It is a structural diagram of a terminal provided in an embodiment of the present application. DETAILED DESCRIPTION

[0052] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0053] This application embodiment provides a method for obtaining map construction data, please refer to Figure 1 , which shows a schematic diagram of an implementation environment provided by an embodiment of the present application. The implementation environment may include: a terminal 11 and a server 12.

[0054] Among them, the terminal 11 is installed with an application capable of acquiring images. When the application acquires an image, the acquired image is sent to the server 12. The server 12 can acquire map construction data based on the method provided in the embodiment of the present application. Optionally, the terminal 11 can acquire map construction data from the server 12, store the map construction data, and generate a map based on the map construction data.

[0055] Alternatively, the terminal 11 is installed with an application capable of acquiring images. When the application acquires the image, the map construction data is acquired based on the method provided in the embodiment of the present application. The terminal 11 can also generate a map based on the map construction data. Optionally, the terminal 11 can send the map construction data to the server 12, and the server 12 stores the map construction data and generates a map based on the map construction data.

[0056] Alternatively, the terminal 11 obtains an image from the server 12 and obtains map construction data based on the method provided in the embodiment of the present application. The terminal 11 can also generate a map based on the map construction data. Optionally, the terminal 11 can send the map construction data to the server 12, and the server 12 stores the map construction data and generates a map based on the map construction data.

[0057] Optionally, the terminal 11 can be any electronic product that can interact with the user through one or more methods such as keyboard, touchpad, touch screen, remote control, voice interaction or handwriting device, such as PC (Personal Computer), mobile phone, smart phone, PDA (Personal Digital Assistant), wearable device, PPC (Pocket PC), tablet computer, smart car machine, smart TV, smart speaker, etc. The server 12 can be a server, a server cluster composed of multiple servers, or a cloud computing service center. The terminal 11 establishes a communication connection with the server 12 through a wired or wireless network.

[0058] Those skilled in the art should understand that the above-mentioned terminal 11 and server 12 are only examples, and other existing or future terminals or servers that are applicable to the present application should also be included in the protection scope of the present application and are included here by reference.

[0059] based on Figure 1 In the implementation environment shown in FIG. 1 , the process of the method for obtaining map construction data provided in the embodiment of the present application is as follows: Figure 2 As shown, the method for acquiring map construction data can be executed by a terminal or a server, and the method includes steps 201 to 204.

[0060] Step 201: Acquire a second label that meets task requirements based on a first label of first image data.

[0061] In a possible implementation, before acquiring the second label that meets the task requirement based on the first label of the first image data, the process further includes acquiring the first image data and determining the label of the first image data.

[0062] The embodiment of the present application does not limit the method of obtaining the first image data. Exemplarily, an open source data set is obtained, and the first image data is obtained from the open source data set. The open source data set refers to a data set that is taken with a high-definition camera under high-quality conditions and is publicly available in a data set related to a road network. For example, the first image data is image data with a resolution of H×W in an open source data set including traffic signs. H represents height and W represents width.

[0063] Optionally, in addition to acquiring the first image data from an open source data set, an image may be acquired by an image acquisition device, and the acquired image may be used as the first image data.

[0064] Regardless of the method used to obtain the first image data, the first image data may have a first label. For example, taking the example of obtaining the first image data from an open source dataset, the open source dataset includes not only the first image data, but also the first label of the first image data. Therefore, the first label of the first image data can be obtained from the open source dataset. Exemplarily, the first label may correspond to different tasks, and the first label may be an image with the content of the task. The embodiment of the present application does not limit the task corresponding to the first label. For example, the task corresponding to the first label is a task such as traffic sign detection, lane line detection, and semantic segmentation. Taking the example of the first label corresponding to the traffic sign detection task, the first label is an image with a traffic sign. Taking the example of the first label corresponding to the lane line detection task, the first label is an image with a lane line. Taking the example of the first label corresponding to the semantic segmentation task, the first label is an image with different elements to be segmented.

[0065] After obtaining the first label, a second label that meets the task requirements can be obtained based on the first label. The embodiment of the present application does not limit the method of obtaining the second label that meets the task requirements based on the first label. The method of obtaining the second label that meets the task requirements based on the first label includes but is not limited to: obtaining a second neural network model based on the first noise and the first label training; inputting the second noise into the second neural network model to obtain a second label that meets the task requirements.

[0066] Optionally, the first noise is a one-dimensional Gaussian noise of length N that conforms to the Gaussian distribution. Wherein N is a positive number. When the second neural network model is obtained based on the first noise and the first label training, an adversarial learning method can be adopted. The training models required in the adversarial learning include a first generative model and a first discriminant model, and the first generative model and the first discriminant model are neural network models.

[0067] Exemplarily, the first generation model and the first discriminant model are CNN (Convolutional Neural Networks) models. The first label is used as a supervisory signal, the first noise is input into the first generation model, and an intermediate label is output according to the first noise, the intermediate label is input into the first discriminant model, and the first discriminant model determines whether the input label is the first label. Alternatively, the first noise is input into the first generation model, and an intermediate label is output according to the first noise, the intermediate label and the first label are randomly input into the first discriminant model, and the first discriminant model determines whether the input label is the first label.

[0068] If the first discriminant model determines that the input label is the first label, it outputs a first value, for example, the first value is 1; if the first discriminant model determines that the input label is the label output by the first generation model, it outputs a second value, for example, the second value is 0.

[0069] During the training process using the above method, the generation loss and the discrimination loss are used to constrain adversarial learning, so that the learning rate of the first generation model is consistent with the learning rate of the first discrimination model. Among them, the generation loss is the difference loss between the intermediate label and the first label, and the discrimination loss is the loss of the first discriminant model judging the first label as false and the intermediate label as true. When the discrimination loss and the generation loss no longer decrease, the training is completed, and the first generation model at the completion of the training is used as the second neural network model.

[0070] Regardless of the method used to obtain the second neural network model, the second noise can be input into the second neural network model to obtain a second label that meets the task requirements. Among them, the first label corresponds to different tasks, and the second neural network model trained using the first labels corresponding to different tasks ultimately outputs labels corresponding to different tasks. For example, taking the corresponding task of traffic sign detection as an example, the first label is an image with a traffic sign, and in the process of training the second neural network model, the first label is used as a supervisory signal, and the generation standard of the first generation model and the discrimination standard of the first discrimination model are both the first label of the supervisory signal. Therefore, the second neural network model obtained by training ultimately outputs a second label that is an image with a traffic sign that matches the first label.

[0071] Optionally, the second noise is a one-dimensional Gaussian noise conforming to the Gaussian distribution and having a length of M. M is a positive number, and M is the same as or different from N. The embodiment of the present application does not limit the values ​​of M and N.

[0072] Optionally, in addition to outputting the second label using the second neural network model, the second label may also be obtained from an open source dataset by acquiring an open source dataset.

[0073] Exemplarily, no matter which method is used to obtain the second label, the second label can correspond to different tasks. The embodiment of the present application does not limit the task corresponding to the second label. For example, the task corresponding to the second label is traffic sign detection, lane line detection, semantic segmentation and other tasks. Optionally, the second label is an image with the content of the task. Taking the second label corresponding to the traffic sign detection task as an example, the second label is an image with a traffic sign. Taking the second label corresponding to the lane line detection task as an example, the second label is an image with lane lines. Taking the second label corresponding to the semantic segmentation task as an example, the second label is an image with different elements to be segmented.

[0074] Step 202: Generate second image data matching the first image data based on the second tag.

[0075] After obtaining the second label, the second image data matching the first image data can be generated based on the second label. The embodiment of the present application does not limit the method of generating the second image data matching the first image data based on the second label. The method of generating the second image data matching the first image data based on the second label includes but is not limited to: training a third neural network model based on the first label and the first image data; inputting the second label into the third neural network model to obtain the second image data matching the first image data.

[0076] When the third neural network model is obtained by training based on the first label and the first image data, an adversarial learning method can be used. The training models required in the adversarial learning include a second generation model and a second discrimination model, and the second generation model and the second discrimination model are neural network models.

[0077] Exemplarily, the second generation model and the second discriminant model use a U-type neural network model, where U is the shape of the neural network. The first image data is used as a supervisory signal, the first label is input into the second generation model, and the intermediate image data is output according to the first label, and the intermediate image data is input into the second discriminant model, and the second discriminant model determines whether the input image data is the first image data. Alternatively, the intermediate image data and the first image data are randomly input into the second discriminant model, and the second discriminant model determines whether the input image data is the first image data.

[0078] If the second discriminant model determines that the input image data is the first image data, it outputs a first value, for example, the first value is 1; if the second discriminant model determines that the input image data is the image data output by the second generation model, it outputs a second value, for example, the second value is 0.

[0079] In the process of training in the above manner, the generation loss and the discrimination loss are used to constrain adversarial learning, so that the learning rate of the second generation model is consistent with the learning rate of the second discrimination model. Among them, the generation loss is the difference loss between the intermediate image data and the first image data, and the discrimination loss is the loss of the second discrimination model judging the first image data as false and the intermediate image data as true. When the discrimination loss and the generation loss no longer decrease, the training is completed, and the second generation model at the completion of the training is used as the third neural network model.

[0080] Regardless of the method used to obtain the third neural network model, the second label can be input into the network model to obtain the second image data that meets the task requirements. In the process of training the third neural network model, the first image data is used as a supervisory signal, and the generation standard of the second generation model and the discrimination standard of the second discrimination model are both the first image data as the supervisory signal, so the final output of the trained third neural network matches the first image data.

[0081] During the adversarial learning training process, the second generation model will be continuously optimized to reduce the difference loss between the image data generated by the second generation model and the first image data, thereby improving the accuracy and similarity of the generated image data.

[0082] Step 203: obtain a first neural network model through training based on the second image data.

[0083] After acquiring the second image data, the first neural network model can be obtained by training based on the second image data. The embodiment of the present application does not limit the training method of obtaining the first neural network model based on the second image data.

[0084] Optionally, the method of obtaining the first neural network model based on the second image data training is: determining the style transfer type; and obtaining the first neural network model based on the second image data and information about the style transfer type training.

[0085] Before determining the type of style transfer, it is also necessary to classify the types of style transfer. Style transfer is to use an algorithm to learn the style of an image and apply the style to another image, changing the style of the image without changing the image content.

[0086] The embodiment of the present application does not limit the classification method of style transfer types. Exemplarily, the map construction data is classified according to the quality style difference as the style transfer type.

[0087] Figure 3 is a quality difference diagram of map construction data provided by an embodiment of the present application, such as Figure 3 As shown, when the map construction data is street view image data, it can be divided into five types according to the quality style differences, namely, "normal", "rainy day", "blurred", "too bright", and "too dark".

[0088] Regardless of the method used to divide the style transfer type, the style transfer type needs to be determined. The embodiment of the present application does not limit the method for determining the style transfer type. The method for determining the style transfer type includes but is not limited to: providing an input box and determining the style transfer type according to the input information; or providing a type option and determining the style transfer type according to the selected option.

[0089] Before the first neural network model is trained based on the second image data and the information of the style transfer type, reference mapping data needs to be obtained. The embodiment of the present application does not limit the method for obtaining the reference mapping data. For example, a driving recorder is obtained and the reference mapping data is collected using the driving recorder.

[0090] When the first neural network model is obtained by training based on the second image data and the information of the style transfer type, an adversarial learning method can be used. The training models required in the adversarial learning include a third generative model and a third discriminative model, and both the third generative model and the third discriminative model use neural network algorithms.

[0091] Exemplarily, the third generative model and the third discriminant model can be CNN models. The reference mapping data is used as a supervisory signal to align and splice the second image data and the information of the style transfer type, and obtain a tensor generated by aligning and splicing the second image data and the information of the style transfer type. The tensor is input into the third generative model, and intermediate image data is generated according to the tensor, and the intermediate image data is input into the third discriminant model, and the third discriminant model determines whether the input image data is the reference mapping data. Alternatively, the tensor is input into the third generative model, and intermediate image data is generated according to the tensor, and the intermediate image data and the reference mapping data are randomly input into the third discriminant model, and the third discriminant model determines whether the input image data is the reference mapping data.

[0092] If the third discriminant model determines that the input image data is reference mapping data, it outputs a first value, for example, the first value is 1; if the third discriminant model determines that the input image data is intermediate image data, it outputs a second value, for example, the second value is 0.

[0093] During the training process using the above method, the generation loss and the discrimination loss are used to constrain adversarial learning, so that the learning rate of the third generation model is consistent with the learning rate of the third discrimination model. Among them, the generation loss is the difference loss between the intermediate image data and the reference mapping data, and the discrimination loss is the loss of the third discrimination model judging the reference mapping data as false and the intermediate image data as true. When the discrimination loss and the generation loss no longer decrease, the training is completed, and the third generation model at the completion of the training is used as the first neural network model.

[0094] The first neural network model is a neural network model that generates controllable data types. The second image data is aligned and spliced ​​with the style transfer type, and the data type variable is determined before input, and then the generated data type is determined, so that the first neural network model can generate controllable data types.

[0095] Optionally, in addition to determining the style transfer type and training the first neural network model based on the second image data and information about the style transfer type, the first neural network model can also be trained based on the first image data and the second image data.

[0096] Before training the first neural network model based on the first image data and the second image data, reference mapping data needs to be obtained. The present application embodiment does not limit the method for obtaining reference mapping data. For example, a driving recorder is obtained and reference mapping data is collected using the driving recorder.

[0097] The first neural network model is obtained by training based on the first image data and the second image data. An adversarial learning method may be adopted. The training models required in the adversarial learning include a fourth generation model and a fourth discrimination model, and the fourth generation model and the fourth discrimination model are neural network models.

[0098] Exemplarily, the fourth generative model and the fourth discriminant model are CNN models. The reference mapping data is used as a supervisory signal, the first image data or the second image data is input into the fourth generative model, and the intermediate image data is generated according to the first image data or the second image data, and the intermediate image data is input into the fourth discriminant model, and the fourth discriminant model determines whether the input image data is the reference mapping data. Alternatively, the first image data or the second image data is input into the fourth generative model, and the intermediate image data is generated according to the first image data or the second image data, and the intermediate image data and the reference mapping data are randomly input into the fourth discriminant model, and the fourth discriminant model determines whether the input image data is the reference mapping data.

[0099] If the fourth discriminant model determines that the input image data is reference mapping data, it outputs a first value, for example, the first value is 1; if the fourth discriminant model determines that the input image data is intermediate image data, it outputs a second value, for example, the second value is 0.

[0100] During the training process using the above method, the generation loss and the discrimination loss are used to constrain adversarial learning, so that the learning rate of the fourth generation model is consistent with the learning rate of the fourth discrimination model. Among them, the generation loss is the difference loss between the intermediate image data and the reference mapping data, and the discrimination loss is the loss of the fourth discrimination model judging the reference mapping data as false and the intermediate image data as true. When the discrimination loss and the generation loss no longer decrease, the training is completed, and the fourth generation model at the completion of the training is used as the first neural network model.

[0101] Optionally, in addition to obtaining the first neural network model by training based on the first image data and the second image data, a style transfer type may also be determined; and the first neural network model may be obtained by training based on information of the first image data, the second image data and the style transfer type.

[0102] Before determining the type of style transfer, it is also necessary to classify the types of style transfer. Style transfer is to use an algorithm to learn the style of an image and apply the style to another image, changing the style of the image while keeping the image content unchanged.

[0103] The embodiment of the present application does not limit the way of dividing the style transfer types. For example, the style transfer types can be divided into "normal", "rainy day", "blurred", "too bright", and "too dark" according to the quality style difference.

[0104] Regardless of the method used to divide the style transfer type, the style transfer type needs to be determined. The embodiment of the present application does not limit the method for determining the style transfer type. The method for determining the style transfer type includes but is not limited to: providing an input box and determining the style transfer type according to the input information; or providing a type option and determining the style transfer type according to the selected option.

[0105] Before obtaining the first neural network model based on the first image data, the second image data and the information of the style transfer type, reference mapping data needs to be obtained. The embodiment of the present application does not limit the method for obtaining the reference mapping data. For example, a driving recorder is obtained and the reference mapping data is collected using the driving recorder.

[0106] When the first neural network model is obtained by training based on the first image data, the second image data and the information of the style transfer type, an adversarial learning method can be used. The training models required in the adversarial learning include the fifth generative model and the fifth discriminative model, and the fifth generative model and the fifth discriminative model are both neural network models.

[0107] Exemplarily, the fifth generative model and the fifth discriminant model are CNN models. The reference mapping data is used as a supervisory signal, and the second image data or the first image data and the information of the style transfer type are aligned and spliced ​​to obtain a tensor generated by the alignment and splicing. The tensor is input into the fifth generative model, and intermediate image data is generated according to the tensor, and the intermediate image data is input into the fifth discriminant model, and the fifth discriminant model determines whether the input image data is the reference mapping data. Alternatively, the tensor is input into the fifth generative model, and intermediate image data is generated according to the tensor, and the intermediate image data and the reference mapping data are randomly input into the fifth discriminant model, and the fifth discriminant model determines whether the input image data is the reference mapping data.

[0108] If the fifth discriminant model determines that the input image data is reference mapping data, it outputs a first value, for example, the first value is 1; if the fifth discriminant model determines that the input image data is intermediate image data, it outputs a second value, for example, the second value is 0.

[0109] During the training process using the above method, the generation loss and the discrimination loss are used to constrain adversarial learning, so that the learning rate of the fifth generation model is consistent with the learning rate of the fifth discrimination model. Among them, the generation loss is the difference loss between the intermediate image data and the reference mapping data, and the discrimination loss is the loss of the fifth discrimination model judging the reference mapping data as false and the intermediate image data as true. When the discrimination loss and the generation loss no longer decrease, the training is completed, and the fifth generation model at the completion of the training is used as the first neural network model.

[0110] Step 204: Perform style migration on the third image data through the first neural network model to obtain map construction data.

[0111] In a possible implementation, before performing style transfer on the third image data through the first neural network model, the method further includes acquiring the third image data.

[0112] The embodiment of the present application does not limit the method of obtaining the third image data. Exemplarily, an open source data set is obtained, and the third image data set is obtained from the open source data set. The open source data set refers to a data set that is taken with a high-definition camera under high-quality conditions and is publicly available in a data set related to a road network. For example, the third image data is image data with a resolution of H×W in an open source data set including traffic signs. H represents height and W represents width.

[0113] Optionally, in addition to acquiring the third image data from the open source data set, an image may be acquired by an image acquisition device, and the acquired image may be used as the third image data.

[0114] Optionally, in addition to acquiring the third image data through the image acquisition device, the third neural network may be acquired, a label corresponding to the task may be input into the third neural network, and the third image data may be output through the third neural network.

[0115] By using the third neural network to output the third image data, the user can generate image data of the corresponding style type in a targeted manner when there is a lack of image data of the corresponding style type in the open source dataset, thereby achieving image content diversification.

[0116] After acquiring the third image data, the third image data and information on the style transfer type are input into the first neural network model, and the style of the third image data is transferred through the first neural network model to obtain map construction data with the style transfer type.

[0117] In summary, the method for obtaining map construction data provided in the embodiment of the present application obtains a second label that meets the task requirements based on the first label of the first image data, and generates second image data that matches the first image data based on the second label. The second image data finally obtained is similar to the first image data. On this basis, the first neural network model is trained using the second image data. Even if the first image data is missing, the second image data can be used for model training to obtain a neural network model that can generate images in multiple styles. During the training process, the input is the tensor generated by aligning and splicing the second image data with the style transfer type. Since the style transfer type is determined before the input, the output style transfer type is controlled, which reduces the training difficulty of the first neural network model and improves the performance of the trained neural network model. Therefore, when the third image data is style-transferred based on the first neural network model, the style type of the obtained map construction data is controllable, the style transfer is natural, and the image content is diverse.

[0118] In one possible implementation, please see Figure 4 , Figure 4 FIG. 1 is a flowchart of obtaining map data provided by an embodiment of the present application. The method is applied to a server as an example for explanation. Figure 4 As shown, when the method for obtaining map construction data is applied to the lane line detection task in a street scene, the method includes the following steps.

[0119] Step 401, input noise to the second neural network model to obtain a label, which is an RGB (Red Green Blue) image with lane lines.

[0120] Step 402 , input the label to the third neural network model, and output image data having a style and content similar to the open source data and containing lane lines.

[0121] Step 403: Fill in the style transfer type information in the input box and determine that the style transfer type is normal.

[0122] In step 404 , align image data with lane lines and style transfer types that are similar in style and content to the open source data, and generate a tensor.

[0123] Step 405: Input the tensor into the first neural network model, and obtain image data whose style transfer type is regular and contains lane lines based on the output result of the first neural network model.

[0124] The image data generated by the above method can effectively improve the performance in the lane detection task. The optimization effect is shown in Table 1:

[0125] Table 1

[0126]

[0127] As can be seen from Table 1, the present application scheme is compared with the reference algorithm from the perspectives of pixel accuracy, pixel false positive rate, pixel false negative rate, region intersection ratio, lane line accuracy, and lane line recall rate. The reference algorithm is a neural network model trained based on image data of an existing open source data set, and the present application scheme is a neural network model trained based on image data with a style and content similar to the open source data and containing lane lines. The data in the "Pixel Accuracy" column in Table 1 above is the quotient of the number of correctly judged pixels and the total number of pixels, the data in the "Pixel False Positive Rate" column is the proportion of pixel samples that are actually false to be judged as true, the data in the "Pixel False Negative Rate" column is the proportion of pixel samples that are actually true to be judged as false, the data in the "Region Intersection Ratio" column is the ratio of the intersection and union of the bounding box judged by the neural network model to the actual bounding box, the data in the "Lane Line Accuracy" column is the number of correctly judged lane lines divided by the total number of lane lines, and the data in the "Lane Line Recall Rate" column is the proportion of all positive samples that are real lane lines in the positive samples.

[0128] Based on the values ​​in Table 1, it can be seen that the neural network model trained by the technical solution provided in the embodiment of the present application has better performance in the lane line detection task.

[0129] In one possible implementation, the method for obtaining map construction data provided by the present application can be applied to the task of isolating belt segmentation in street scene. Schematically, noise is input to the second neural network model to obtain a label, which is an image containing an isolating belt. The label is input to the third neural network model to obtain image data containing an isolating belt whose style and content are similar to the open source data; fill in the style transfer type in the input box, determine that the style transfer type is blurred, align and splice the image data with similar style and content to the open source data and containing the isolating belt and the style transfer type, and generate a tensor; input the tensor to the first neural network model to obtain blurred image data containing the isolating belt.

[0130] The image data generated by the above method can effectively improve the performance in the isolation zone segmentation task. The optimization effects are shown in Table 2 and Table 3:

[0131] Table 2

[0132]

[0133] As can be seen from Table 2, the present application scheme is compared with the reference algorithm from the perspectives of accuracy, recall, category average area intersection-over-union ratio, background area intersection-over-union ratio, vegetation area intersection-over-union ratio, and fence area intersection-over-union ratio. The reference algorithm is a neural network model trained based on image data of an existing open source data set, while the present application scheme is a neural network model trained based on image data with a style and content similar to the open source data and containing isolation zones. In Table 2 above, the data in the "Accuracy" column is the quotient of the number of correctly judged isolation zones and the total number of isolation zones, the data in the "Recall" column is the proportion of the real isolation zones in the positive samples to all positive samples, the data in the "Category Average Area Intersection-Union Ratio" column is the ratio of the intersection and union of the category average area judged by the neural network model to the actual category average area, the data in the "Background Area Intersection-Union Ratio" column is the ratio of the intersection and union of the background area judged by the neural network model to the actual background area, the data in the "Vegetation Area Intersection-Union Ratio" column is the ratio of the intersection and union of the vegetation area judged by the neural network model to the actual vegetation area, and the data in the "Fence Area Intersection-Union Ratio" column is the ratio of the intersection and union of the fence area judged by the neural network model to the actual fence area.

[0134] Based on the numerical values ​​in Table 2, it can be seen that the neural network model obtained based on the technical solution provided in this application can improve the accuracy, recall rate, category average area intersection-over-union ratio, background area intersection-over-union ratio, vegetation area intersection-over-union ratio and fence area intersection-over-union ratio of the isolation belt segmentation task, and has better performance in the isolation belt segmentation task.

[0135] Table 3

[0136]

[0137] As can be seen from Table 3, the present application scheme is compared with the reference algorithm from the perspectives of the intersection-and-joint ratio of the pier area, the intersection-and-joint ratio of the sidewalk area, the intersection-and-joint ratio of the vegetation guardrail area, the intersection-and-joint ratio of the non-vegetation guardrail area, the intersection-and-joint ratio of the highway area, and the intersection-and-joint ratio of other areas. The reference algorithm is a neural network model trained based on image data of an existing open source data set, while the present application scheme is a neural network model trained based on image data with a style and content similar to the open source data and containing isolation belts. In the above Table 3, the data in the column "Intersection-Union Ratio of Pier Area" is the ratio of the intersection and union of the pier area judged by the neural network model to the actual pier area, the data in the column "Intersection-Union Ratio of Sidewalk Area" is the ratio of the intersection and union of the sidewalk area judged by the neural network model to the actual sidewalk area, the data in the column "Intersection-Union Ratio of Vegetation Guardrail Area" is the ratio of the intersection and union of the vegetation guardrail area judged by the neural network model to the actual vegetation guardrail area, the data in the column "Intersection-Union Ratio of Non-Vegetation Guardrail Area" is the ratio of the intersection and union of the non-vegetation guardrail area judged by the neural network model to the actual non-vegetation guardrail area, the data in the column "Intersection-Union Ratio of Highway Area" is the ratio of the intersection and union of the highway area judged by the neural network model to the actual highway area, and the data in the column "Intersection-Union Ratio of Other Areas" is the ratio of the intersection and union of other areas judged by the neural network model to the actual other areas.

[0138] Based on the numerical values ​​in Table 3, it can be seen that the neural network model obtained based on the technical solution provided in this application can improve the intersection-and-merge ratio of the pier area, the intersection-and-merge ratio of the sidewalk area, the intersection-and-merge ratio of the vegetation guardrail area, the intersection-and-merge ratio of the non-vegetation guardrail area, the intersection-and-merge ratio of the highway area and the intersection-and-merge ratio of other areas, and has better performance in the isolation belt segmentation task.

[0139] See also Figure 5 , an embodiment of the present application provides a structural schematic diagram of a device for obtaining map construction data, the device comprising: an acquisition module 501, a data generation module 502, a training module 503, and a style transfer module 504.

[0140] An acquisition module 501 is used to acquire a second label that meets the task requirements based on the first label of the first image data;

[0141] A data generating module 502, configured to generate second image data matching the first image data based on the second tag;

[0142] A training module 503, configured to obtain a first neural network model through training based on the second image data;

[0143] The style transfer module 504 is used to perform style transfer on the third image data through the first neural network model to obtain map construction data.

[0144] Optionally, the training module 503 is used to determine the style transfer type; and obtain the first neural network model through training based on the second image data and information about the style transfer type.

[0145] Optionally, the style transfer module 504 is used to input the third image data and information about the style transfer type into the first neural network model, perform style transfer on the third image data through the first neural network model, and obtain map construction data with the style transfer type.

[0146] Optionally, the acquisition module 501 is used to obtain a first label of the first image data; train a second neural network model based on the first noise and the first label; and input a second noise into the second neural network model to obtain a second label that meets the task requirements.

[0147] Optionally, the data generation module 502 is used to obtain first image data; train a third neural network model based on the first label and the first image data; input the second label to the third neural network model, and output the second image data through the third neural network model.

[0148] Optionally, the training module 503 is used to train a first neural network model based on the first image data and the second image data.

[0149] Optionally, the training module 503 is used to determine the style transfer type; and obtain a first neural network model through training based on the first image data, the second image data and information about the style transfer type.

[0150] The above-mentioned device obtains a second label that meets the task requirements, and generates second image data matching the first image data based on the second label. Since the second label is obtained based on the first label of the first image data, and the second image data matches the first image data, the second image data can be made similar to the first image data. On this basis, the second image data is used to train the first neural network model. Even if the first image data is missing, the generated second image data can still be used for model training, which improves the performance of the trained neural network model. Therefore, when the style of the third image data is transferred based on the first neural network model, the image quality of the obtained map construction data is higher and the accuracy is improved.

[0151] It should be noted that the device provided in the above embodiment only uses the division of the above functional modules as an example to implement its functions. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and their specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0152] Figure 6 This is a schematic diagram of the structure of a server provided in an embodiment of the present application. The server may have relatively large differences due to different configurations or performances, and may include one or more processors (Central Processing Units, CPU) 601 and one or more memories 602, wherein the one or more memories 602 store at least one computer program, and the at least one computer program is loaded and executed by the one or more processors 601, so that the server implements the method for obtaining map construction data provided in the above-mentioned various method embodiments. Of course, the server may also have components such as a wired or wireless network interface, a keyboard, and an input and output interface for input and output. The server may also include other components for implementing device functions, which will not be repeated here.

[0153] Figure 7 1 is a schematic diagram of a terminal structure provided by an embodiment of the present application. For example, the terminal is a smart phone, a tablet computer, an MP3 (Moving Picture Experts Group Audio Layer III) player, an MP4 (Moving Picture Experts Group Audio Layer IV) player, a laptop computer or a desktop computer. The terminal may also be referred to as a user device, a portable terminal, a laptop terminal, a desktop terminal or other names.

[0154] Typically, the terminal includes: a processor 701 and a memory 702 .

[0155] The processor 701 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 701 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 701 may also include a main processor and a coprocessor. The main processor is a processor for processing data in an awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in a standby state. In some embodiments, the processor 701 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 701 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.

[0156] The memory 702 may include one or more computer-readable storage media, which may be non-transitory. The memory 702 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 702 is used to store at least one instruction, which is used to be executed by the processor 701 so that the terminal implements the method for obtaining map construction data provided in the method embodiment of the present application.

[0157] In some embodiments, the terminal may further optionally include: a peripheral device interface 703 and at least one peripheral device. The processor 701, the memory 702 and the peripheral device interface 703 may be connected via a bus or a signal line. Each peripheral device may be connected to the peripheral device interface 703 via a bus, a signal line or a circuit board. Specifically, the peripheral device includes: at least one of a radio frequency circuit 704, a display screen 705, a camera assembly 706, an audio circuit 707, a positioning assembly 708 and a power supply 709.

[0158] The peripheral device interface 703 may be used to connect at least one peripheral device related to I / O (Input / Output) to the processor 701 and the memory 702. In some embodiments, the processor 701, the memory 702, and the peripheral device interface 703 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 701, the memory 702, and the peripheral device interface 703 may be implemented on a separate chip or circuit board, which is not limited in this embodiment.

[0159] The radio frequency circuit 704 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 704 communicates with communication networks and other communication devices through electromagnetic signals. The radio frequency circuit 704 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals into electrical signals. Optionally, the radio frequency circuit 704 includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a user identity module card, and the like. The radio frequency circuit 704 can communicate with other terminals through at least one wireless communication protocol. The wireless communication protocol includes, but is not limited to: a metropolitan area network, various generations of mobile communication networks (2G, 3G, 4G and 5G), a wireless local area network and / or a WiFi (Wireless Fidelity) network. In some embodiments, the radio frequency circuit 704 may also include circuits related to NFC (Near Field Communication), which is not limited in this application.

[0160] The display screen 705 is used to display the UI (User Interface). The UI may include graphics, text, icons, videos and any combination thereof. When the display screen 705 is a touch display screen, the display screen 705 also has the ability to collect touch signals on the surface or above the surface of the display screen 705. The touch signal can be input to the processor 701 as a control signal for processing. At this time, the display screen 705 can also be used to provide virtual buttons and / or virtual keyboards, also known as soft buttons and / or soft keyboards. In some embodiments, the display screen 705 can be one, set on the front panel of the terminal; in other embodiments, the display screen 705 can be at least two, respectively set on different surfaces of the terminal or in a folding design; in other embodiments, the display screen 705 can be a flexible display screen, set on a curved surface or a folding surface of the terminal. Even, the display screen 705 can also be set to a non-rectangular irregular shape, that is, a special-shaped screen. The display screen 705 can be made of materials such as LCD (Liquid Crystal Display), OLED (Organic Light-Emitting Diode) and the like.

[0161] The camera assembly 706 is used to capture images or videos. Optionally, the camera assembly 706 includes a front camera and a rear camera. Typically, the front camera is arranged on the front panel of the terminal, and the rear camera is arranged on the back of the terminal. In some embodiments, there are at least two rear cameras, which are any one of a main camera, a depth of field camera, a wide-angle camera, and a telephoto camera, so as to realize the fusion of the main camera and the depth of field camera to realize the background blur function, the fusion of the main camera and the wide-angle camera to realize panoramic shooting and VR (Virtual Reality) shooting function or other fusion shooting functions. In some embodiments, the camera assembly 706 may also include a flash. The flash can be a monochrome temperature flash or a dual-color temperature flash. A dual-color temperature flash refers to a combination of a warm light flash and a cold light flash, which can be used for light compensation at different color temperatures.

[0162] The audio circuit 707 may include a microphone and a speaker. The microphone is used to collect sound waves from the user and the environment, and convert the sound waves into electrical signals and input them into the processor 701 for processing, or input them into the radio frequency circuit 704 to achieve voice communication. For the purpose of stereo acquisition or noise reduction, there may be multiple microphones, which are respectively arranged at different parts of the terminal. The microphone may also be an array microphone or an omnidirectional acquisition microphone. The speaker is used to convert the electrical signal from the processor 701 or the radio frequency circuit 704 into sound waves. The speaker may be a traditional film speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, it can not only convert the electrical signal into sound waves audible to humans, but also convert the electrical signal into sound waves inaudible to humans for purposes such as ranging. In some embodiments, the audio circuit 707 may also include a headphone jack.

[0163] Positioning component 708 is used to locate the current geographical location of the terminal to implement navigation or LBS (Location Based Service). Positioning component 708 can be a positioning component based on the US GPS (Global Positioning System), China's Beidou system, Russia's Grenas system or the European Union's Galileo system.

[0164] The power supply 709 is used to power various components in the terminal. The power supply 709 can be an alternating current, a direct current, a disposable battery, or a rechargeable battery. When the power supply 709 includes a rechargeable battery, the rechargeable battery can support wired charging or wireless charging. The rechargeable battery can also be used to support fast charging technology.

[0165] In some embodiments, the terminal further includes one or more sensors 710 , including but not limited to: an acceleration sensor 711 , a gyroscope sensor 712 , a pressure sensor 713 , a fingerprint sensor 714 , an optical sensor 715 , and a proximity sensor 716 .

[0166] The acceleration sensor 711 can detect the magnitude of acceleration on the three coordinate axes of the coordinate system established by the terminal. For example, the acceleration sensor 711 can be used to detect the components of gravity acceleration on the three coordinate axes. The processor 701 can control the display screen 705 to display the user interface in a horizontal view or a vertical view according to the gravity acceleration signal collected by the acceleration sensor 711. The acceleration sensor 711 can also be used for collecting game or user motion data.

[0167] The gyroscope sensor 712 can detect the body direction and rotation angle of the terminal, and the gyroscope sensor 712 can cooperate with the acceleration sensor 711 to collect the user's 3D actions on the terminal. The processor 701 can implement the following functions based on the data collected by the gyroscope sensor 712: motion sensing (such as changing the UI according to the user's tilt operation), image stabilization during shooting, game control, and inertial navigation.

[0168] The pressure sensor 713 can be set in the side frame of the terminal and / or the lower layer of the display screen 705. When the pressure sensor 713 is set in the side frame of the terminal, the user's holding signal of the terminal can be detected, and the processor 701 performs left and right hand recognition or shortcut operation according to the holding signal collected by the pressure sensor 713. When the pressure sensor 713 is set in the lower layer of the display screen 705, the processor 701 controls the operability controls on the UI interface according to the user's pressure operation on the display screen 705. The operability controls include at least one of a button control, a scroll bar control, an icon control, and a menu control.

[0169] The fingerprint sensor 714 is used to collect the user's fingerprint, and the processor 701 identifies the user's identity based on the fingerprint collected by the fingerprint sensor 714, or the fingerprint sensor 714 identifies the user's identity based on the collected fingerprint. When the user's identity is identified as a trusted identity, the processor 701 authorizes the user to perform relevant sensitive operations, which include unlocking the screen, viewing encrypted information, downloading software, paying, and changing settings. The fingerprint sensor 714 can be set on the front, back or side of the terminal. When a physical button or a manufacturer logo is set on the terminal, the fingerprint sensor 714 can be integrated with the physical button or the manufacturer logo.

[0170] The optical sensor 715 is used to collect the ambient light intensity. In one embodiment, the processor 701 can control the display brightness of the display screen 705 according to the ambient light intensity collected by the optical sensor 715. Specifically, when the ambient light intensity is high, the display brightness of the display screen 705 is increased; when the ambient light intensity is low, the display brightness of the display screen 705 is reduced. In another embodiment, the processor 701 can also dynamically adjust the shooting parameters of the camera assembly 706 according to the ambient light intensity collected by the optical sensor 715.

[0171] The proximity sensor 716, also known as a distance sensor, is usually arranged on the front panel of the terminal. The proximity sensor 716 is used to collect the distance between the user and the front of the terminal. In one embodiment, when the proximity sensor 716 detects that the distance between the user and the front of the terminal is gradually decreasing, the processor 701 controls the display screen 705 to switch from the screen-on state to the screen-off state; when the proximity sensor 716 detects that the distance between the user and the front of the terminal is gradually increasing, the processor 701 controls the display screen 705 to switch from the screen-off state to the screen-on state.

[0172] Those skilled in the art will understand that Figure 7 The structure shown in the figure does not constitute a limitation on the terminal, and may include more or less components than those shown in the figure, or combine certain components, or adopt a different component arrangement.

[0173] In an exemplary embodiment, a computer device is also provided, the computer device comprising a processor and a memory, wherein at least one computer program is stored in the memory. The at least one computer program is loaded and executed by one or more processors, so that the computer device implements any of the above methods for obtaining map construction data.

[0174] In an exemplary embodiment, a computer-readable storage medium is also provided, in which at least one computer program is stored. The at least one computer program is loaded and executed by a processor of a computer device to enable the computer to implement any of the above methods for obtaining map construction data.

[0175] In one possible implementation, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, etc.

[0176] In an exemplary embodiment, a computer program product or a computer program is also provided, the computer program product or the computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs any of the above methods for obtaining map construction data.

[0177] It should be understood that the "plurality" mentioned in this article refers to two or more. "And / or" describes the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects are in an "or" relationship.

[0178] The above description is only an exemplary embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for obtaining map construction data, characterized in that: The method comprises: Acquire a second label that meets the task requirements based on the first label of the first image data; generating second image data matching the first image data based on the second tag; A first neural network model is obtained by training based on the second image data; The third image data is style-transferred by the first neural network model to obtain map construction data; and the second label that meets the task requirements is obtained based on the first label of the first image data, including: A second neural network model is obtained by training based on the first noise and the first label; Inputting a second noise into the second neural network model to obtain a second label that meets task requirements; The generating second image data matching the first image data based on the second tag comprises: acquiring the first image data; A third neural network model is obtained by training based on the first label and the first image data; The second label is input into the third neural network model, and the second image data is output through the third neural network model.

2. The method according to claim 1, characterized in that The step of obtaining a first neural network model based on the second image data training comprises: Determine the style transfer type; A first neural network model is obtained by training based on the second image data and the information of the style transfer type.

3. The method according to claim 2, characterized in that The performing style migration on the third image data by using the first neural network model to obtain map construction data includes: The third image data and information about the style transfer type are input into the first neural network model, and the style of the third image data is transferred through the first neural network model to obtain map construction data with the style transfer type.

4. The method according to claim 1, characterized in that The step of obtaining a first neural network model based on the second image data training comprises: A first neural network model is obtained by training based on the first image data and the second image data.

5. The method according to claim 1, characterized in that The step of obtaining a first neural network model based on the second image data training comprises: Determine the style transfer type; A first neural network model is obtained by training based on the first image data, the second image data and information of the style transfer type.

6. A device for obtaining map construction data, characterized in that: The device comprises: An acquisition module, used for acquiring a second label that meets the task requirements based on the first label of the first image data; a data generating module, configured to generate second image data matching the first image data based on the second tag; A training module, used for training a first neural network model based on the second image data; A style transfer module, used to perform style transfer on the third image data through the first neural network model to obtain map construction data; The acquiring of a second label that meets the task requirement based on the first label of the first image data includes: A second neural network model is obtained by training based on the first noise and the first label; Inputting a second noise into the second neural network model to obtain a second label that meets task requirements; The generating second image data matching the first image data based on the second tag comprises: acquiring the first image data; A third neural network model is obtained by training based on the first label and the first image data; The second label is input into the third neural network model, and the second image data is output through the third neural network model.

7. A computer device, characterized in that: The computer device includes a processor and a memory, wherein at least one computer program is stored in the memory, and the at least one computer program is loaded and executed by the processor so that the computer device implements the method for obtaining map construction data as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one computer program, and the at least one computer program is loaded and executed by a processor so that the computer implements the method for obtaining map construction data as described in any one of claims 1 to 5.

9. A computer program product, characterized in that The computer program product comprises a computer program or instructions, and the computer program or instructions are executed by a processor to enable a computer to implement the method for obtaining map construction data according to any one of claims 1 to 5.

Images