Data processing method and device, detection method and device, and equipment

Abstract

Embodiments of the present application provide a data processing method and device, a detection method and device, and equipment. A target detection model is constructed according to a house door detection task and a door point detection task. A sample house panoramic image and label data are obtained. The label data includes at least one house door object label and a plurality of target point labels on the at least one house door object. The target detection model is trained by using the sample house panoramic image and the label data. The target detection model is used to detect a house panoramic image to be processed, and at least one house door object and a plurality of target points on the at least one house door object are obtained. The technical solution provided by the embodiments of the present application improves the detection effect of the target points on the house door object.

Description

Technical Field

[0001] This application relates to the field of image processing, and more particularly to a data processing method and apparatus, a detection method and apparatus, and a device. Background Technology

[0002] In the field of housing-related business, it is common practice to detect target points on doors in panoramic images of houses. These target points can include, for example, the points where the door intersects with the ground. This allows for panoramic image stitching based on the target points to obtain the overall layout and configuration of the house. Traditional solutions typically require first identifying the door objects in the panoramic image, and then detecting the target points on the door objects. This process is cumbersome, time-consuming, and ineffective. Summary of the Invention

[0003] This application provides a data processing method and apparatus, a detection method and apparatus, and a device to solve the problem of poor target point detection effect on door objects in the prior art.

[0004] In a first aspect, this application provides a data processing method, including:

[0005] Based on the door detection task and the door point detection task, construct a target detection model;

[0006] Acquire panoramic images and annotation data of sample houses; the annotation data includes at least one door object label and multiple target point labels on at least one door object.

[0007] The target detection model is trained using the sample panoramic images of houses and the labeled data; wherein the target detection model is used to detect the panoramic images of houses to be processed, and obtain at least one door object and multiple target points on the at least one door object.

[0008] Optionally, constructing the target detection model based on the door detection task and the door point detection task includes:

[0009] The input module, intermediate processing module, and output modules corresponding to the door detection task and door point detection task are built sequentially to obtain the target detection model.

[0010] Optionally, training the target detection model using the sample panoramic images of houses and the labeled data includes:

[0011] The sample house panoramic image is input into the target detection model to identify at least one predicted door object and multiple predicted target points on the at least one predicted door object;

[0012] The target detection model is trained using the at least one predicted door object and multiple predicted target points on the at least one predicted door object, along with the labeled data.

[0013] Optionally, the annotation data includes a first vertex coordinate label and a second vertex coordinate label in the outer detection frame of at least one door object, and a first target point coordinate label and a second target point coordinate label on the at least one door object that intersects with the ground;

[0014] The step of inputting the panoramic image of the sample house into the target detection model to identify at least one predicted door object and multiple predicted target points on the at least one predicted door object includes:

[0015] The sample house panoramic image is input into the target detection model to identify and obtain the first predicted vertex coordinates and the second predicted vertex coordinates of the outer detection box of at least one door object, as well as the first predicted target point coordinates and the second predicted target point coordinates of the at least one door object that intersects with the ground.

[0016] The step of training the target detection model using the at least one predicted door object and multiple predicted target points on the at least one predicted door object, along with the labeled data, includes:

[0017] Based on the difference information between the first predicted vertex coordinates and the first vertex coordinate label, the difference information between the second predicted vertex coordinates and the second vertex coordinate label, the difference information between the first predicted target point coordinates and the first target point coordinate label, and the difference information between the second predicted target point coordinates and the second target point coordinate label, the loss value is calculated;

[0018] The model parameters of the target detection model are optimized using the loss value until the loss value meets the loss condition, thereby obtaining the target detection model.

[0019] Secondly, this application provides a detection method, including:

[0020] Obtain a panoramic image of the house to be processed;

[0021] The panoramic image of the house is input into the target detection model to identify at least one door object and multiple target points on the at least one door object;

[0022] The target detection model is trained based on panoramic images of sample houses and labeled data. The labeled data includes at least one door object label and multiple target point labels on at least one door object.

[0023] Optionally, inputting the panoramic image of the house into the target detection model to identify at least one door object and multiple target points on the at least one door object includes:

[0024] The panoramic image of the house is input into the target detection model to identify and obtain the coordinates of the first vertex and the second vertex of the outer detection frame of at least one door object, as well as the coordinates of the first target point and the second target point on the at least one door object that intersects with the ground.

[0025] Optionally, the method further includes:

[0026] Based on at least one door object corresponding to each of the multiple panoramic images of houses to be processed, and multiple target points on the at least one door object, the multiple panoramic images of houses to be processed are stitched together.

[0027] Thirdly, this application provides a data processing apparatus, comprising:

[0028] The building module is used to construct the target detection model based on the door detection task and the door point detection task;

[0029] The first acquisition module is used to acquire panoramic images and annotation data of sample houses; the annotation data includes at least one door object label and multiple target point labels on at least one door object.

[0030] The training module is used to train the target detection model using the sample panoramic images of houses and the labeled data; wherein the target detection model is used to detect the panoramic images of houses to be processed, and obtain at least one door object and multiple target points on the at least one door object.

[0031] Fourthly, this application provides a detection device, comprising:

[0032] The second acquisition module is used to acquire panoramic images of the houses to be processed.

[0033] The detection module is used to input the panoramic image of the house into the target detection model, identify at least one door object, and multiple target points on the at least one door object;

[0034] The target detection model is trained based on panoramic images of sample houses and labeled data. The labeled data includes at least one door object label and multiple target point labels on at least one door object.

[0035] Fifthly, this application provides an electronic device, including a storage component and a processing component;

[0036] The storage component stores one or more computer instructions, which are invoked and executed by the processing component; the processing component invokes the one or more computer instructions to implement the data processing method as described in the first aspect, or the detection method as described in the second aspect.

[0037] In a sixth aspect, this application provides a computer-readable storage medium storing a computer program, which, when executed by a computer, implements the data processing method as described in the first aspect or the detection method as described in the second aspect.

[0038] In this embodiment, a target detection model can be constructed based on the door detection task and the door point detection task, and sample panoramic images of houses and labeled data can be obtained. The labeled data may include at least one door object label and multiple target point labels on at least one door object. The target detection model is trained using the sample panoramic images of houses and labeled data. The trained target detection model can detect the panoramic images of houses to be processed, and obtain at least one door object and multiple target points on at least one door object. That is, door object detection and target point detection on door objects can be completed simultaneously, without the need to first identify door objects and then detect target points on door objects, simplifying the detection process, shortening the detection time, and improving detection efficiency.

[0039] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 A flowchart of one embodiment of a data processing method provided in this application is shown;

[0042] Figure 2 A flowchart of one embodiment of the detection method provided in this application is shown;

[0043] Figure 3 This invention provides a schematic diagram of the structure of one embodiment of a data processing apparatus.

[0044] Figure 4 This invention provides a schematic diagram of the structure of one embodiment of a detection device.

[0045] Figure 5 A schematic diagram of the structure of an embodiment of an electronic device provided in this application is shown. Detailed Implementation

[0046] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0047] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0048] The technical solution of this application is applicable to the field of image processing, especially in image processing in business scenarios related to housing, such as panoramic image processing of houses. In the business field related to housing, target points on the doors in the panoramic image corresponding to the house can usually be detected. Target points may include, for example, the points where the door intersects with the ground, so as to obtain the overall layout and configuration of the house by panoramic image stitching based on the target points.

[0049] Traditional solutions typically require first identifying door objects in the panoramic image, and then detecting target points on the door objects. This process is cumbersome, time-consuming, and ineffective.

[0050] To address the aforementioned technical problems, the inventors, after a series of considerations and experiments, proposed the technical solution of this application, providing a data processing method, including: constructing a target detection model based on a door detection task and a door point detection task; acquiring a sample house panoramic image and annotation data; the annotation data including at least one door object label and multiple target point labels on the at least one door object; training the target detection model using the sample house panoramic image and the annotation data; wherein, the target detection model is used to detect the house panoramic image to be processed, and obtain at least one door object and multiple target points on the at least one door object.

[0051] In this embodiment, a target detection model can be constructed based on the door detection task and the door point detection task, and sample panoramic images of houses and labeled data can be obtained. The labeled data may include at least one door object label and multiple target point labels on at least one door object. The target detection model is trained using the sample panoramic images of houses and labeled data. The trained target detection model can detect the panoramic images of houses to be processed, and obtain at least one door object and multiple target points on at least one door object. That is, door object detection and target point detection on door objects can be completed simultaneously, without the need to first identify door objects and then detect target points on door objects, simplifying the detection process, shortening the detection time, and improving detection efficiency.

[0052] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0053] The technical solutions of this application embodiment can be applied to a system architecture including a user terminal and a server terminal, where the user terminal and the server terminal establish a connection through a network. The network provides a medium for the communication link between the user terminal and the server terminal. The network can include various connection types, such as wired or wireless communication links or fiber optic cables, etc.

[0054] The user client can interact with the server via the network to receive or send messages, etc.

[0055] The user end can be a browser, an app (application), a web application such as an H5 (HyperText Markup Language 5) application, a mini-program (also known as a lightweight application), or a cloud application. The user end can be deployed on electronic devices and depends on the device to run or on certain apps within the device. Electronic devices can have displays and support information browsing, such as personal mobile terminals like smartphones, tablets, and personal computers. Various other types of applications can also be configured on electronic devices, such as human-computer interaction applications, model training applications, text processing applications, web browser applications, shopping applications, search applications, instant messaging tools, email clients, and social media platforms.

[0056] The server side can include servers that provide various services, such as servers that support the models used on the user's end for background training, or servers that process interactive information sent by the user's end.

[0057] It should be noted that the server can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. The server can also be a server in a distributed system, or a server combined with blockchain. The server can also be a cloud server, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology.

[0058] It should be noted that the data processing and detection methods provided in the embodiments of this application are generally executed by the server, and the corresponding target detection model is generally set in the server. However, in other embodiments of this application, the user terminal may also have similar functions to the server, thereby executing the data processing and detection methods provided in the embodiments of this application. In other embodiments, the data processing and detection methods provided in the embodiments of this application may also be jointly executed by the user terminal and the server.

[0059] The technical solution of this application embodiment is implemented through a target detection model. The model generation process is briefly described below. Figure 1 The diagram shown is a flowchart of an embodiment of a data processing method provided in this application. The method may include the following steps.

[0060] S11: Construct a target detection model based on the door detection task and the target point detection task.

[0061] S12: Obtain panoramic images and annotation data of the sample houses.

[0062] The labeled data may include at least one door object label and multiple target point labels on at least one door object.

[0063] S13: Train the target detection model using panoramic images of sample houses and labeled data.

[0064] The target detection model can be used to detect panoramic images of houses to be processed, and obtain at least one door object and multiple target points on the door object.

[0065] This embodiment is applicable to door detection and target point detection on doors in panoramic images of a house. A panoramic image of a house can refer to an isometric cylindrical projection of the house. A panoramic image typically includes one or more door objects, and each door can include one or more target points. The target points can be set according to the actual application scenario, such as the center point of the door, corner points, or the point where the door frame intersects with the ground, without limitation.

[0066] Based on the tasks of door detection and target point detection, a target detection model can be constructed. This target detection model can employ neural network models, such as Convolutional Neural Networks (CNN), Fully Convolutional Networks (FCN), U-Net (a U-shaped structure, an improved FCN structure), and U-Net++ (composed of U-Nets of varying depths).

[0067] In an optional embodiment, based on the door detection task and the target point detection task, an input module, an intermediate processing module, and output modules corresponding to the door detection task and the target point detection task can be sequentially constructed to obtain a target detection model. Specifically, the output module can simultaneously output the detection information of at least one door object, such as the position coordinates of the bounding box of the door object, and the detection information of one or more target points on each door object, such as the position coordinates of the target points.

[0068] Specifically, a large number of sample house panoramic images and annotation data can be obtained. The annotation data can include at least one door object label in the sample house panoramic image, such as the position coordinates of the outer detection box of the door object in the sample house panoramic image, and multiple target point labels on at least one door object, such as the position coordinates of the target points on the door object in the sample house panoramic image.

[0069] Then, the target detection model can be trained using the sample house panoramic images and labeled data. Specifically, the sample house panoramic images can be input into the target detection model to identify at least one predicted door object and multiple predicted target points on the predicted door object. The target detection model can then be trained using the at least one predicted door object, the multiple predicted target points on the predicted door object, and the labeled data.

[0070] In this embodiment, a target detection model can be constructed based on the door detection task and the door point detection task, and sample panoramic images of houses and labeled data can be obtained. The labeled data can include at least one door object label and multiple target point labels on at least one door object. The target detection model is trained using the sample panoramic images of houses and labeled data. The trained target detection model can detect the panoramic images of houses to be processed, and obtain at least one door object and multiple target points on at least one door object. That is, it can simultaneously complete the detection of both door objects and target points on door objects, without having to first identify door objects and then detect target points on door objects, simplifying the detection process, shortening the detection time, and improving detection efficiency.

[0071] In an optional embodiment, the above-mentioned annotation data may include coordinate labels for a first vertex and a second vertex of at least one detection frame of a door object, and coordinate labels for a first target point and a second target point on the door object that intersects with the ground. Optionally, the first vertex may refer to the upper left corner of the detection frame of the door object, and the second vertex may refer to the lower right corner of the detection frame of the door object. The first target point and the second target point may refer to any two points on the door object that intersect with the ground.

[0072] At this point, the method of inputting the panoramic image of the sample house into the target detection model to identify at least one predicted door object and multiple predicted target points on the at least one predicted door object may include:

[0073] The panoramic image of the sample house is input into the target detection model to identify the coordinates of the first and second predicted vertices in the bounding box of at least one door object, as well as the coordinates of the first and second predicted target points on at least one door object that intersect with the ground. The first predicted vertex can refer to the upper left corner of the bounding box of the predicted door object, and the second predicted target point can refer to the lower right corner of the bounding box of the predicted door object.

[0074] Furthermore, methods for training an object detection model using at least one predicted door object and multiple predicted target points and labeled data on at least one predicted door object may include:

[0075] Based on the differences between the first predicted vertex coordinates and the first vertex coordinate label, the differences between the second predicted vertex coordinates and the second vertex coordinate label, the differences between the first predicted target point coordinates and the first target point coordinate label, and the differences between the second predicted target point coordinates and the second target point coordinate label, the loss value is calculated.

[0076] The model parameters of the target detection model are optimized using the loss value until the loss value meets the loss condition, thus obtaining the target detection model.

[0077] Specifically, the loss value is calculated using the predicted data and labeled data, and the loss value is used to propagate the model in the direction of the error. The specific implementation of optimizing the model parameters can refer to the implementation process in the traditional solution, and will not be elaborated here.

[0078] Considering that both the detection of door objects and the detection of target points on door objects utilize features from the vicinity of the door, this paper proposes a method to combine door object detection and target point detection. By using features from door detection to complete target point detection, two detection tasks can be performed simultaneously. Optimizing features through both tasks promotes mutual improvement, enhancing the model's detection accuracy, recall, and precision across both tasks.

[0079] The technical solution of this application will be briefly described below from the perspective of image detection.

[0080] like Figure 2 The diagram shown is a flowchart of an embodiment of a detection method provided in this application. The method may include the following steps.

[0081] S21: Obtain the panoramic image of the house to be processed.

[0082] S22: Input the panoramic image of the house into the target detection model to identify at least one door object and multiple target points on the door object.

[0083] The target detection model is trained based on panoramic images of sample houses and labeled data. The labeled data may include at least one door object label and multiple target point labels on at least one door object.

[0084] In this embodiment, the panoramic image of the house to be processed may include one or more door objects, and each door object may include one or more target points. Using a pre-trained object detection model, at least one door object and multiple target points on that door object can be identified. The training process of the object detection model can be referred to... Figure 1 The relevant descriptions of the embodiments shown will not be repeated here.

[0085] In this embodiment, the target detection model can be used to detect the panoramic image of the house to be processed, and obtain at least one door object and multiple target points on the door object. That is, the detection of the door object and the target points on the door object can be completed simultaneously. There is no need to first identify the door object and then detect the target points on the door object, which simplifies the detection process, shortens the detection time, and improves the detection efficiency.

[0086] In an optional embodiment, a method for inputting a panoramic image of a house into a target detection model to identify at least one door object and multiple target points on the at least one door object may include:

[0087] The panoramic image of the house is input into the object detection model to identify the coordinates of the first and second vertices of the bounding box of at least one door object, as well as the coordinates of the first and second target points on the door object that intersect with the ground. Optionally, the first vertex can refer to the top-left corner of the bounding box of the door object, and the second vertex can refer to the bottom-right corner of the bounding box of the door object. The first and second target points can refer to any two points on the door object that intersect with the ground.

[0088] In practical applications, after detecting and obtaining the door object and target points on the door in the panoramic image, the panoramic image can be stitched together using the door object and target points to obtain a floor plan of the house, etc. Therefore, in an optional embodiment, the above method may further include:

[0089] Based on at least one door object corresponding to each of the multiple panoramic images of houses to be processed, and multiple target points on the at least one door object, the multiple panoramic images of houses to be processed are stitched together.

[0090] Specifically, the implementation of stitching based on door objects and target points on door objects in multiple panoramic images can be set according to the actual application scenario. For example, two panoramic images containing the same door object can be stitched together, etc., without any restrictions.

[0091] Figure 3 This is a schematic diagram of the structure of one embodiment of a data processing device provided in this application. The device may include the following modules.

[0092] Module 301 is used to build a target detection model based on the door detection task and the target point detection task;

[0093] The first acquisition module 302 is used to acquire panoramic images and annotation data of sample houses; the annotation data includes at least one door object label and multiple target point labels on at least one door object.

[0094] Training module 303 is used to train the target detection model using the sample panoramic image of the house and the labeled data; wherein, the target detection model is used to detect the panoramic image of the house to be processed and obtain at least one door object and multiple target points on the at least one door object.

[0095] In an optional embodiment, the construction module 301 can be specifically used to build an input module, an intermediate layer processing module, and output modules corresponding to the door detection task and the target point detection task in sequence according to the door detection task and the target point detection task, so as to obtain a target detection model.

[0096] In an optional embodiment, the training module 303 may be specifically used to input the sample house panoramic image into the target detection model, identify at least one predicted door object and multiple predicted target points on the at least one predicted door object; and train the target detection model using the at least one predicted door object and the multiple predicted target points on the at least one predicted door object and the labeled data.

[0097] In an optional embodiment, the annotation data includes a first vertex coordinate label and a second vertex coordinate label in the outer detection frame of at least one door object, and a first target point coordinate label and a second target point coordinate label on the at least one door object that intersects with the ground.

[0098] The training module 303 can be specifically used to input the panoramic image of the sample house into the target detection model, identify and obtain the first predicted vertex coordinates and the second predicted vertex coordinates of the outer detection box of at least one door object, as well as the first predicted target point coordinates and the second predicted target point coordinates intersecting with the ground on the at least one door object; calculate the loss value based on the difference information between the first predicted vertex coordinates and the first vertex coordinate label, the difference information between the second predicted vertex coordinates and the second vertex coordinate label, the difference information between the first predicted target point coordinates and the first target point coordinate label, and the difference information between the second predicted target point coordinates and the second target point coordinate label; and optimize the model parameters of the target detection model using the loss value until the loss value meets the loss condition to obtain the target detection model.

[0099] Figure 4 This is a schematic diagram of one embodiment of a detection device provided in this application. The device may include the following modules.

[0100] The second acquisition module 401 is used to acquire a panoramic image of the house to be processed;

[0101] The detection module 402 is used to input the panoramic image of the house into the target detection model, identify at least one door object, and multiple target points on the at least one door object;

[0102] The target detection model is trained based on panoramic images of sample houses and labeled data. The labeled data includes at least one door object label and multiple target point labels on at least one door object.

[0103] In an optional embodiment, the detection module 402 can be specifically used to input the panoramic image of the house into the target detection model, identify and obtain the coordinates of the first vertex and the second vertex of the outer detection frame of at least one door object, as well as the coordinates of the first target point and the second target point on the at least one door object that intersects with the ground.

[0104] In an optional embodiment, the device may further include:

[0105] The stitching module is used to stitch together the multiple panoramic images of houses to be processed, based on at least one door object corresponding to each of the multiple panoramic images of houses to be processed and multiple target points on the at least one door object.

[0106] This application also provides an electronic device, such as... Figure 5 As shown, the device may include a storage component 501 and a processing component 502.

[0107] The storage component 501 stores one or more computer program instructions, wherein one or more computer program instructions are invoked and executed by the processing component 502 to achieve... Figure 1 The data processing method shown, or Figure 2 The detection method shown.

[0108] The processing component 502 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0109] Storage component 501 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0110] Of course, the aforementioned electronic devices may also include other components, such as input / output interfaces, communication components, etc.

[0111] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc. Communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0112] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The data processing method of the illustrated embodiment, or Figure 2 The detection method of the illustrated embodiment. The computer-readable medium may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device.

[0113] This application also provides a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program, when executed by a computer, can perform the above-described functions. Figure 1 The data processing method of the illustrated embodiment, or Figure 2 The detection method of the embodiment shown.

[0114] In such embodiments, the computer program may be downloaded and installed from a network, and / or installed from a removable medium. When the computer program is executed by a processor, it performs the various functions defined in the system of this application.

[0115] It should be noted that the aforementioned computing devices can be physical devices or elastic computing hosts provided by cloud computing platforms. They can be implemented as a distributed cluster of multiple servers or terminal devices, or as a single server or a single terminal device.

[0116] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0117] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0118] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0119] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A data processing method, characterized in that, include: Based on the door detection task and the target point detection task, construct a target detection model; Obtain panoramic images and annotation data of sample houses; The labeled data includes the first vertex coordinate label and the second vertex coordinate label of the outer detection frame of at least one door object corresponding to the panoramic image of the sample house, as well as the first target point coordinate label and the second target point coordinate label of the at least one door object intersecting with the ground. The sample house panoramic image is input into the target detection model to identify and obtain the first predicted vertex coordinates and the second predicted vertex coordinates of the outer detection box of at least one door object, as well as the first predicted target point coordinates and the second predicted target point coordinates of the at least one door object that intersects with the ground. Based on the difference information between the first predicted vertex coordinates and the first vertex coordinate label, the difference information between the second predicted vertex coordinates and the second vertex coordinate label, the difference information between the first predicted target point coordinates and the first target point coordinate label, and the difference information between the second predicted target point coordinates and the second target point coordinate label, the loss value is calculated; The model parameters of the target detection model are optimized using the loss value until the loss value meets the loss condition to obtain the target detection model; wherein, the target detection model is used to detect the panoramic image of the house to be processed, and obtain the coordinates of the first vertex and the second vertex of the outer detection box of at least one door object, as well as the coordinates of the first target point and the second target point on the at least one door object that intersects with the ground.

2. The method according to claim 1, characterized in that, The construction of the target detection model based on the door detection task and the target point detection task includes: Based on the door detection task and the target point detection task, the input module, the intermediate layer processing module, and the output modules corresponding to the door detection task and the target point detection task are built in sequence to obtain the target detection model.

3. A detection method, characterized in that, include: Obtain a panoramic image of the house to be processed; The panoramic image of the house is input into the target detection model to identify at least one door object and multiple target points on the at least one door object; The step of inputting the panoramic image of the house into the target detection model to identify at least one door object and multiple target points on the at least one door object includes: The panoramic image of the house is input into the target detection model to identify and obtain the coordinates of the first vertex and the second vertex of the outer detection box of at least one door object, as well as the coordinates of the first target point and the second target point on the at least one door object that intersects with the ground. The target detection model receives a panoramic image of a sample house as input and identifies the first and second predicted vertex coordinates within the outer detection frame of at least one door object, as well as the first and second predicted target point coordinates intersecting the ground on the at least one door object. Based on the differences between the first and second predicted vertex coordinates and their labels, the second and third predicted vertex coordinates and their labels, the first and second predicted target point coordinates and their labels, and the second and third predicted target point coordinates and their labels, the model calculates a loss value and uses the loss value to optimize the model parameters until the loss value meets the loss condition, thus training the model.

4. The method according to claim 3, characterized in that, The method further includes: Based on at least one door object corresponding to each of the multiple panoramic images of houses to be processed, and multiple target points on the at least one door object, the multiple panoramic images of houses to be processed are stitched together.

5. A data processing apparatus, characterized in that, include: The building module is used to construct the target detection model based on the door detection task and the door point detection task; The first acquisition module is used to acquire panoramic images and annotation data of sample houses; The labeled data includes the first vertex coordinate label and the second vertex coordinate label of the outer detection frame of at least one door object corresponding to the panoramic image of the sample house, as well as the first target point coordinate label and the second target point coordinate label of the at least one door object intersecting with the ground. The training module is used to input the panoramic image of the sample house into the target detection model, identify and obtain the first predicted vertex coordinates and the second predicted vertex coordinates of the outer detection box of at least one door object, as well as the first predicted target point coordinates and the second predicted target point coordinates of the at least one door object that intersects with the ground. Based on the difference information between the first predicted vertex coordinates and the first vertex coordinate label, the difference information between the second predicted vertex coordinates and the second vertex coordinate label, the difference information between the first predicted target point coordinates and the first target point coordinate label, and the difference information between the second predicted target point coordinates and the second target point coordinate label, the loss value is calculated; The model parameters of the target detection model are optimized using the loss value until the loss value meets the loss condition to obtain the target detection model; wherein, the target detection model is used to detect the panoramic image of the house to be processed, and obtain the coordinates of the first vertex and the second vertex of the outer detection box of at least one door object, as well as the coordinates of the first target point and the second target point on the at least one door object that intersects with the ground.

6. A detection device, characterized in that, include: The second acquisition module is used to acquire panoramic images of the houses to be processed. The detection module is used to input the panoramic image of the house into the target detection model, identify and obtain the coordinates of the first vertex and the second vertex of the outer detection frame of at least one door object, as well as the coordinates of the first target point and the second target point on the at least one door object that intersects with the ground; The target detection model receives a panoramic image of a sample house as input and identifies the first and second predicted vertex coordinates within the outer detection frame of at least one door object, as well as the first and second predicted target point coordinates intersecting the ground on the at least one door object. Based on the differences between the first and second predicted vertex coordinates and their labels, the second and third predicted vertex coordinates and their labels, the first and second predicted target point coordinates and their labels, and the second and third predicted target point coordinates and their labels, the model calculates a loss value and uses the loss value to optimize the model parameters until the loss value meets the loss condition, thus training the model.

7. An electronic device, characterized in that, Includes storage components and processing components; The storage component stores one or more computer instructions, which are invoked and executed by the processing component; the processing component invokes the one or more computer instructions to implement the data processing method as described in any one of claims 1 to 2, or the detection method as described in any one of claims 3 to 4.

8. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a computer, implements the data processing method as described in any one of claims 1 to 2, or the detection method as described in any one of claims 3 to 4.

Images