Marking processing method, device and equipment, and storage medium
By generating three views of furniture and performing image preprocessing, and using a pre-trained classification and labeling model for automated labeling, the problems of high cost and unstable accuracy in existing technologies are solved, and low-cost, high-accuracy automated identification of furniture data labels is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- GUANGDONG SANWEIJIA INFORMATION TECH CO LTD
- Filing Date
- 2022-08-31
- Publication Date
- 2026-06-02
AI Technical Summary
Existing marking technologies suffer from high costs and unstable accuracy, especially when furniture data formats are inconsistent. Manual marking consumes a lot of manpower, while automated recognition has low accuracy.
By acquiring the 3D model data of the target object, generating three views and performing image preprocessing, the compressed image data is then input into a pre-trained classification and labeling model for classification and labeling. CNN technology is used to realize an automated model labeling process.
It has achieved an automated model labeling process, which reduces labor costs while ensuring recognition accuracy, and solves the problems of unstable accuracy and high cost in existing technologies.
Smart Images
Figure CN115457316B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of marking technology, and in particular to a marking processing method, apparatus, equipment and storage medium. Background Technology
[0002] Currently, labeling is an essential operational step for most companies involved in data exchange across the industry chain. Only when data formats are consistent can other workflows be automated. However, apart from a few data points with universally accepted standard formats, even if the stored data content is largely the same, most data formats differ between companies. For example, furniture manufacturers lack a universal, standardized labeling system. When importing 3D models provided by furniture manufacturers into design platforms for users to choose from, issues such as missing or conflicting labels frequently arise, requiring significant manpower for manual data cleaning and labeling to standardize the design platform's specifications.
[0003] Specifically, for home design platforms, manual labeling and directly inputting furniture images into Convolutional Neural Networks (CNNs) for classification and label output are the most common labeling methods. However, both methods have their drawbacks. Manual labeling requires a large number of professionals, and when dealing with complex labels such as style tags that require specialized knowledge, different people often obtain different labels, consuming significant time and effort, resulting in high costs and unstable label accuracy. While directly inputting furniture images into a CNN to output labels reduces labor costs, for certain types of furniture, a single, single-angle image is insufficient to determine specific attributes. For example, a top-down view is sufficient for a CNN model to learn various feature labels for a bed, but a front view is more effective for a chandelier, and vice versa. Furthermore, due to the different structural features of different furniture types, there is no single optimal angle, making it impossible to determine accurate labels for all objects based on images from fixed angles, leading to unstable labeling accuracy. Summary of the Invention
[0004] In view of this, this application provides a marking processing method, apparatus, equipment and storage medium to realize an automated model marking process, thereby solving the problems of high cost and unstable accuracy in existing marking technologies.
[0005] Firstly, this application provides a marking processing method, including:
[0006] Obtain the 3D model data of the target object;
[0007] Based on the three-dimensional model data, generate three views of the target object;
[0008] Image preprocessing is performed based on the three views to obtain compressed image data;
[0009] The compressed image data is input into a pre-trained classification and labeling model for classification and labeling to obtain the label information corresponding to the target object.
[0010] Optionally, generating three views of the target object based on the three-dimensional model data includes:
[0011] Based on the three-dimensional model data, image rendering is performed to obtain a three-dimensional model view of the target object;
[0012] The front view, side view, and top view corresponding to the three-dimensional model view are defined as the three views.
[0013] Optionally, the step of performing image preprocessing based on the three views to obtain image compressed data includes:
[0014] Determine the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view respectively;
[0015] Image compression data is obtained by performing color channel concatenation processing based on the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view.
[0016] Optionally, determining the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view respectively includes:
[0017] The front view is separated according to color channel information to obtain first color channel image data, second color channel image data and third color channel image data, and the first color channel image data, second color channel image data and third color channel image data are used as the color channel image data of the front view;
[0018] The side view is separated according to the color channel information to obtain the fourth color channel image data, the fifth color channel image data and the sixth color channel image data, and the fourth color channel image data, the fifth color channel image data and the sixth color channel image data are used as the color channel image data of the side view;
[0019] The top view is separated according to the color channel information to obtain the seventh color channel image data, the eighth color channel image data, and the ninth color channel image data. The seventh color channel image data, the eighth color channel image data, and the ninth color channel image data are used as the color channel image data of the top view.
[0020] Optionally, the color channel concatenation processing based on the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view to obtain image compressed data includes:
[0021] According to a preset channel connection method, the image data of the first color channel, the second color channel, the third color channel, the fourth color channel, the fifth color channel, the sixth color channel, the seventh color channel, the eighth color channel, and the ninth color channel are connected and combined to obtain the image compressed data.
[0022] Optionally, before inputting the compressed data information into the pre-trained model to obtain the label information, the following steps are included:
[0023] Obtain the three views of the sample object and the sample label information corresponding to the sample object;
[0024] Based on the three views of the sample object, image preprocessing is performed to obtain the sample image compressed data corresponding to the sample object;
[0025] The model is trained using the compressed data of the sample images and the sample label information to obtain the classification and labeling model.
[0026] Optionally, the target object includes furniture; the label information includes at least one of style label information, color label information, material label information, and category label information.
[0027] Secondly, this application provides a marking processing apparatus, comprising:
[0028] The 3D model data acquisition module is used to acquire the 3D model data of the target object.
[0029] The three-view determination module is used to generate three views of the target object based on the three-dimensional model data.
[0030] The image compression data determination module is used to perform image preprocessing based on the three views to obtain image compression data.
[0031] The label information determination module is used to input the image compressed data information into a pre-trained classification and labeling model for classification and labeling processing to obtain the label information corresponding to the target object.
[0032] Thirdly, this application provides an electronic device, characterized in that it includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;
[0033] Memory, used to store computer programs;
[0034] When a processor executes a program stored in memory, it implements the steps of the marking processing method as described in any embodiment of the first aspect.
[0035] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the marking processing method as described in any embodiment of the first aspect.
[0036] This application embodiment generates three-view images of the target object based on the obtained three-dimensional model data of the target object, performs image preprocessing based on the three-view images to obtain compressed image data, and then inputs the compressed image data information into a pre-trained classification and labeling model for classification and labeling processing to obtain the label information corresponding to the target object. This realizes an automated model labeling process, solves the problems of unstable accuracy and high cost in existing labeling technologies, and achieves the goal of reducing labor costs while ensuring recognition accuracy. Attached Figure Description
[0037] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0039] Figure 1 A flowchart illustrating a marking process provided in an embodiment of this application;
[0040] Figure 2 A flowchart illustrating a marking process provided in an optional embodiment of this application;
[0041] Figure 3 A schematic diagram of the three views of the target object provided as an example of this application;
[0042] Figure 4 A flowchart illustrating the classification and labeling model training method provided in this application embodiment;
[0043] Figure 5 This is a schematic diagram of a marking processing device provided in an embodiment of this application. Detailed Implementation
[0044] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, 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, 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.
[0045] When users design spaces and select furniture on interior design platforms, the furniture tag filtering function can quickly narrow down their choices, improving the user experience. For example, a user might filter by selecting tags such as (modern, wood, coffee-colored, double bed). In related technologies, considering that different types of furniture have different structural features and no single optimal angle, it's difficult to accurately determine their attributes from a single image. Directly inputting the furniture's 3D model into a CNN model to train the discriminant model can improve tag recognition accuracy. However, this technology is still immature and requires significantly more computing power than CNN models that use images as input.
[0046] One of the core concepts of this application is to propose a labeling processing method that automates the model labeling process by reducing the dimensionality of 3D model data and applying CNN technology, thereby reducing labor costs while ensuring recognition accuracy.
[0047] Specifically, in this embodiment, after obtaining the 3D model data of the target object, a three-view drawing of the target object is generated based on the obtained 3D model data. The three-view drawing is then preprocessed to obtain compressed image data. Subsequently, the compressed image data is input into a pre-trained classification and labeling model for classification and labeling processing. This reduces the dimensionality of the 3D model of the home and uses CNN technology to obtain the label information corresponding to the target object. This achieves an automated model labeling process, solving the problems of unstable accuracy and high cost in existing labeling technologies. It reduces labor costs while ensuring recognition accuracy.
[0048] To facilitate understanding of the embodiments of this application, the following will provide further explanation and description with reference to the accompanying drawings and specific embodiments. These embodiments do not constitute a limitation on the embodiments of this application.
[0049] Figure 1 This is a flowchart illustrating a marking process provided in an embodiment of this application.
[0050] Step 110: Obtain the 3D model data of the target object;
[0051] The target object can refer to the object to be marked, which may include furniture to be marked or other objects; the three-dimensional model data can refer to the model data of a three-dimensional model (object, OBJ) format file, and this application embodiment does not impose specific limitations on this.
[0052] Step 120: Generate three views of the target object based on the three-dimensional model data;
[0053] Specifically, in this embodiment of the application, after obtaining the three-dimensional model data of the target object, the three-dimensional model data can be rendered to obtain a three-dimensional view that can be rotated and moved. Then, the three-dimensional view can be rotated along the x-axis, y-axis and z-axis to obtain the three views of the target object, so that image preprocessing can be performed based on the three views to obtain image compressed data, i.e., step 130 is executed.
[0054] Step 130: Perform image preprocessing based on the three views to obtain compressed image data;
[0055] Specifically, in this embodiment of the application, after obtaining the three views of the target object, the three views are separated according to color channels to obtain the color channel image data corresponding to the front view, top view, and side view respectively. Then, the color channel image data corresponding to the front view, top view, and side view can be combined to determine the image compression data of the target object. In order to input the image compression data into the pre-trained classification and labeling model for classification and labeling processing, the label information corresponding to the target object is obtained, i.e., step 140 is executed.
[0056] The three views can include the front view, top view, and side view of the target object; the color channel data can refer to the image data obtained after separating the three views of the target object according to the color channels. Each color channel data can contain the data information contained in one color channel of an image. The front view, top view, and side view of the target object correspond to three color channel data respectively; the image compression data can refer to the compressed data obtained after combining the color channel data corresponding to the three views of the target object. Specifically, it can refer to the width information, length information, and number of color channels of the three view image of the target object. This application embodiment does not impose specific limitations on this.
[0057] Step 140: Input the compressed image data into a pre-trained classification and labeling model for classification and labeling processing to obtain the label information corresponding to the target object.
[0058] Specifically, in this embodiment of the application, after determining the image compression data of the target object, the image compression data can be input into a pre-trained classification and labeling model to classify and label the target object, thereby obtaining the label information corresponding to the target object. This realizes an automated model labeling process, solves the problems of unstable accuracy and high cost in existing labeling technologies, and achieves the goal of reducing labor costs while ensuring recognition accuracy.
[0059] The label information may include style label information, category label information, color label information, and / or material label information of the target object, etc., and this application embodiment does not impose specific limitations on this. For example, if the target object includes furniture, the label information of a certain piece of furniture may be modern-wood-coffee-double bed.
[0060] As can be seen, the embodiments of this application generate three-view images of the target object based on the obtained three-dimensional model data of the target object, and perform image preprocessing based on the three-view images to obtain image compressed data. Then, the image compressed data information is input into a pre-trained classification and labeling model for classification and labeling processing to obtain the label information corresponding to the target object, thereby realizing an automated model labeling process. This solves the problems of unstable accuracy and high cost in existing labeling technologies, and achieves the goal of reducing labor costs while ensuring recognition accuracy.
[0061] Reference Figure 2 This illustration shows a marking processing method provided by an optional embodiment of this application. Specifically, the marking processing method provided by the optional embodiment of this application may include the following steps:
[0062] Step 210: Obtain the 3D model data of the target object;
[0063] Step 220: Render the image based on the three-dimensional model data to obtain a three-dimensional model view of the target object;
[0064] Step 230: Determine the front view, side view, and top view corresponding to the three-dimensional model view as the three views;
[0065] Specifically, such as Figure 3As shown in the embodiment of this application, after obtaining the three-dimensional model data of the target object, the three-dimensional model data can be rendered by a rendering tool to obtain a three-dimensional model view of the target object. Then, the three-dimensional model view can be rotated 90 degrees clockwise along the x-axis, y-axis, and z-axis respectively, and the rotated view can be cropped to determine the front view, side view, and top view of the target object. Subsequently, the front view, side view, and top view can be determined as the three views of the target object so that image compression data can be determined based on the three views. The image compression data is then input into a pre-trained model to determine the label information of the target object.
[0066] As an example of this application, after obtaining the three-dimensional model data of the target object, the example of this application can input the three-dimensional model data into a three-dimensional (3D) rendering tool, such as 3D MAX, and then render the 3D model data on the computer to visualize the structure and color of the three-dimensional model data, and obtain a rotatable and movable three-dimensional model view corresponding to the three-dimensional model data. Then, the center point of the three-dimensional model view of the target object can be placed at the center point of the computer screen, and the three-dimensional model view can be rotated 90 degrees clockwise along the x-axis, y-axis, and z-axis respectively, and the front view, side view, and top view of uniform size and center alignment can be cropped. The front view, side view, and top view corresponding to the three-dimensional model are determined as the three views, and then the three images are output.
[0067] In addition to capturing the three views of a target object's 3D view, it can also capture other angles of the 3D view, such as a 45-degree oblique view.
[0068] Step 240: Determine the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view, respectively;
[0069] Step 250: Perform color channel concatenation processing based on the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view to obtain image compressed data;
[0070] Step 260: Input the compressed image data into a pre-trained classification and labeling model for classification and labeling processing to obtain the label information corresponding to the target object.
[0071] Specifically, in this embodiment of the application, after determining the three views of the target object, the three views can be processed by color channel separation according to the color channel information to obtain the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view. Then, the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view can be combined to obtain compressed data containing the color channel image data of the three views of the target object. The compressed image data can then be input into a pre-trained classification and labeling model for classification and labeling to obtain the label information corresponding to the target object.
[0072] Optionally, in embodiments of this application, the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view are determined respectively. Specifically, this may include: separating the front view according to color channel information to obtain first color channel image data, second color channel image data, and third color channel image data, and using the first color channel image data, second color channel image data, and third color channel image data as the color channel image data of the front view; separating the side view according to the color channel information to obtain fourth color channel image data, fifth color channel image data, and sixth color channel image data, and using the fourth color channel image data, fifth color channel image data, and sixth color channel image data as the color channel image data of the side view; and separating the top view according to the color channel information to obtain seventh color channel image data, eighth color channel image data, and ninth color channel image data, and using the seventh color channel image data, eighth color channel image data, and ninth color channel image data as the color channel image data of the top view.
[0073] In practical processing, in this embodiment of the application, when obtaining the three views of a target object with an image length and width of 100, each pixel of the image is represented by three color channels (RGB). Therefore, the total dimensions of the front view, top view, and side view of the target object are all 100*100*3. Subsequently, the three views of the target object can be separated according to color channel information to obtain the first color channel image data corresponding to the first color channel of the front view, the second color channel image data corresponding to the second color channel of the front view, and the third color channel image data corresponding to the third color channel of the front view. The first color channel image data, the second color channel image data, and the third color channel image data are then used as the... The color channel image data of the view; the fourth color channel image data corresponding to the first color channel of the side view, the fifth color channel image data corresponding to the second color channel of the side view, and the sixth color channel image data corresponding to the third color channel of the side view, and the fourth color channel image data, the fifth color channel image data, and the sixth color channel image data are used as the color channel image data of the side view; the seventh color channel image data corresponding to the first color channel of the attached view, the eighth color channel image data corresponding to the second color channel of the attached view, and the ninth color channel image data corresponding to the third color channel of the attached view, and the seventh color channel image data, the eighth color channel image data, and the ninth color channel image data are used as the color channel image data of the attached view.
[0074] The first color channel can be the red channel, the second color channel can be the green channel, and the third color channel can be the blue channel. This application does not impose specific limitations on this.
[0075] Optionally, in this embodiment of the application, color channel image data of the front view, color channel image data of the side view, and color channel image data of the top view are processed by color channel concatenation to obtain image compressed data. Specifically, this may include: concatenating and combining the first color channel image data, the second color channel image data, the third color channel image data, the fourth color channel image data, the fifth color channel image data, the sixth color channel image data, the seventh color channel image data, the eighth color channel image data, and the ninth color channel image data according to a preset channel concatenation method to obtain the image compressed data.
[0076] The preset channel connection method can refer to connecting the image data from the first color channel to the ninth color channel in sequence. Of course, in addition to connecting and combining the first, second, third, fourth, fifth, sixth, seventh, eighth, and ninth color channel image data according to the preset channel connection method, this application embodiment can also use other methods to connect the first, second, third, fourth, fifth, sixth, seventh, eighth, and ninth color channel image data. The data can be combined in several ways. For example, based on the first color information, the image data of the first color channel, the image data of the fourth color channel, and the image data of the seventh color channel can be concatenated to obtain compressed data of the first color channel; based on the second color information, the image data of the second color channel, the image data of the fifth color channel, and the image data of the eighth color channel can be concatenated to obtain compressed data of the second color channel; based on the third color information, the image data of the third color channel, the image data of the sixth color channel, and the image data of the ninth color channel can be concatenated to obtain compressed data of the third color channel; and the compressed data of the first color channel, the compressed data of the second color channel, and the compressed data of the third color channel can be combined to obtain compressed image data.
[0077] Specifically, in this embodiment, after obtaining the first, second, and third color channel image data corresponding to the front view of the target object, the fourth, fifth, and sixth color channel image data corresponding to the side view, and the sixth, seventh, eighth, and ninth color channel image data corresponding to the top view, the first, second, third, fourth, fifth, sixth, seventh, eighth, and ninth color channel image data can be processed by channel connection according to a preset channel connection method to obtain image compressed data containing nine color channels. The total dimension of the image compressed data is 100*100*9, so that the label information can be determined based on the image compressed data. This realizes an automated model labeling process, solves the problems of unstable accuracy and high cost in existing labeling technologies, and achieves the goal of reducing labor costs while ensuring recognition accuracy.
[0078] As can be seen, the embodiments of this application generate three-view images of the target object based on the obtained three-dimensional model data of the target object, and perform image preprocessing based on the three-view images to obtain image compressed data. Then, the image compressed data information is input into a pre-trained classification and labeling model for classification and labeling processing to obtain the label information corresponding to the target object, thereby realizing an automated model labeling process. This solves the problems of unstable accuracy and high cost in existing labeling technologies, and achieves the goal of reducing labor costs while ensuring recognition accuracy.
[0079] Reference Figure 4 This illustration shows a labeling processing method provided by an optional embodiment of this application. Specifically, the labeling processing method provided by the optional embodiment of this application can be applied to the model training stage, and the labeling processing method may specifically include the following steps:
[0080] Step 410: Obtain the three views of the sample object and the sample label information corresponding to the sample object;
[0081] Step 420: Perform image preprocessing based on the three views of the sample object to obtain the sample image compressed data corresponding to the sample object;
[0082] Step 430: The model is trained using the compressed sample image data and the sample label information to obtain the classification and labeling model.
[0083] Specifically, in this embodiment of the application, before inputting the compressed data information into a pre-trained model to obtain the label information, the three views of the sample object and the label information of the sample object can be obtained. For example, the label information of a sample object can be obtained as modern-wood-coffee-double bed. Then, the three views of the sample object can be pre-processed to obtain sample image compressed data with a total dimension of image length * image width * 9. The sample image compressed data and sample label information are then input into the deep learning model for model training to obtain a classification and labeling model.
[0084] The classification and labeling models can include style classification and labeling models, category classification and labeling models, color classification and labeling models, and material classification and labeling models. During model training, the three views of a sample object and its style label (e.g., modern style) can be input into the deep learning model to obtain a style classification and labeling model; the three views of a sample object and its category label (e.g., double bed) can be input into the deep learning model to obtain a category classification and labeling model; the three views of a sample object and its material label (e.g., wood) can be input into the deep learning model to obtain a material classification and labeling model; and the three views of a sample object and its color label (e.g., coffee color) can be input into the deep learning model to obtain a color classification and labeling model.
[0085] The deep learning model may include a CNN. Furthermore, the images input to a CNN are typically images with 3 color channels, i.e., compressed image data with a total dimension of image length * image width * 3. However, in this embodiment, the image data input to the CNN is an image with 9 color channels, i.e., compressed image data with a total dimension of image length * image width * 9. Therefore, when inputting the compressed sample image data and sample label information into the CNN, the input channels of the first layer neurons of the CNN need to be adjusted from 3 channels to 9 channels.
[0086] Furthermore, this application also provides a schematic diagram of the structure of a marking processing device, as shown in the embodiment. Figure 5 As shown, the marking processing apparatus 500 provided in this application embodiment may include the following modules:
[0087] The 3D model data acquisition module 510 is used to acquire the 3D model data of the target object; the three-view determination module 520 is used to generate the three-views of the target object based on the 3D model data.
[0088] The image compression data determination module 530 is used to perform image preprocessing based on the three views to obtain image compression data.
[0089] The label information determination module 540 is used to input the image compressed data information into a pre-trained classification and labeling model for classification and labeling processing to obtain the label information corresponding to the target object.
[0090] Optionally, the three-view determination module 520 includes the following sub-modules:
[0091] 3D Model View Determination Submodule: Used to perform image rendering based on the 3D model data to obtain a 3D model view of the target object;
[0092] The three-view determination submodule is used to determine the front view, side view, and top view corresponding to the three-dimensional model view as the three-view diagram.
[0093] Optionally, the image compression data determination module 530 includes the following sub-modules:
[0094] Color channel image data determination submodule: used to determine the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view respectively;
[0095] Image compression data determination submodule: used to perform color channel concatenation processing based on the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view to obtain image compression data.
[0096] Optionally, the color channel image data determination submodule includes the following sub-units:
[0097] The front view color channel image data determination subunit is used to separate the front view according to color channel information to obtain first color channel image data, second color channel image data and third color channel image data, and use the first color channel image data, the second color channel image data and the third color channel image data as the color channel image data of the front view.
[0098] The side view color channel image data determination subunit is used to separate the side view according to the color channel information to obtain fourth color channel image data, fifth color channel image data and sixth color channel image data, and use the fourth color channel image data, fifth color channel image data and sixth color channel image data as the color channel image data of the side view;
[0099] The attached view color channel image data determination subunit is used to separate the top view according to the color channel information to obtain the seventh color channel image data, the eighth color channel image data and the ninth color channel image data, and use the seventh color channel image data, the eighth color channel image data and the ninth color channel image data as the color channel image data of the top view.
[0100] Optionally, the image compression data determination submodule includes the following sub-units:
[0101] The image compression data determination subunit is used to connect and combine the first color channel image data, the second color channel image data, the third color channel image data, the fourth color channel image data, the fifth color channel image data, the sixth color channel image data, the seventh color channel image data, the eighth color channel image data, and the ninth color channel image data according to a preset channel connection method to obtain the image compression data.
[0102] Optionally, the marking processing device 500 further includes:
[0103] The sample object information determination module is used to obtain the three views of the sample object and the sample label information corresponding to the sample object;
[0104] The sample image compressed data determination module is used to perform image preprocessing based on the three views of the sample object to obtain the sample image compressed data corresponding to the sample object.
[0105] The classification and labeling model training module is used to train the model using the compressed data of the sample images and the sample label information to obtain the classification and labeling model.
[0106] Optionally, the target object in this application embodiment includes furniture; the label information includes at least one of style label information, color label information, material label information, and category label information.
[0107] Furthermore, this application embodiment also provides an electronic device, characterized in that it includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory is used to store computer programs; and the processor is used to implement the steps of the marking processing method described in any of the above method embodiments when executing the program stored in the memory.
[0108] Furthermore, embodiments of this application also provide a computer-readable storage medium storing a computer program thereon, characterized in that the computer program, when executed by a processor, implements the steps of the marking processing method described in any of the above method embodiments.
[0109] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0110] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.
Claims
1. A marking processing method, characterized in that, include: Obtain the 3D model data of the target object; Based on the three-dimensional model data, generate three views of the target object; Image preprocessing is performed based on the three views to obtain compressed image data; The compressed image data is input into a pre-trained classification and labeling model for classification and labeling to obtain the label information corresponding to the target object. The step of generating three views of the target object based on the three-dimensional model data includes: Based on the three-dimensional model data, image rendering is performed to obtain a three-dimensional model view of the target object; The front view, side view, and top view corresponding to the three-dimensional model view are defined as the three views; The step of preprocessing the image based on the three views to obtain compressed image data includes: Determine the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view respectively; Image compression data is obtained by performing color channel concatenation processing based on the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view.
2. The method according to claim 1, characterized in that, The process of determining the color channel image data of the front view, the side view, and the top view includes: The front view is separated according to color channel information to obtain first color channel image data, second color channel image data and third color channel image data, and the first color channel image data, second color channel image data and third color channel image data are used as the color channel image data of the front view; The side view is separated according to the color channel information to obtain the fourth color channel image data, the fifth color channel image data and the sixth color channel image data, and the fourth color channel image data, the fifth color channel image data and the sixth color channel image data are used as the color channel image data of the side view; The top view is separated according to the color channel information to obtain the seventh color channel image data, the eighth color channel image data, and the ninth color channel image data. The seventh color channel image data, the eighth color channel image data, and the ninth color channel image data are used as the color channel image data of the top view.
3. The method according to claim 2, characterized in that, The image data is obtained by performing color channel concatenation processing on the color channel image data of the front view, the side view, and the top view, to obtain compressed image data, including: According to a preset channel connection method, the image data of the first color channel, the second color channel, the third color channel, the fourth color channel, the fifth color channel, the sixth color channel, the seventh color channel, the eighth color channel, and the ninth color channel are connected and combined to obtain the image compressed data.
4. The method according to claim 1, characterized in that, Before inputting the compressed data information into the pre-trained model to obtain the label information, the process includes: Obtain the three views of the sample object and the sample label information corresponding to the sample object; Based on the three views of the sample object, image preprocessing is performed to obtain the sample image compressed data corresponding to the sample object; The model is trained using the compressed data of the sample images and the sample label information to obtain the classification and labeling model.
5. The method according to claim 1, characterized in that, The target object includes furniture; the label information includes at least one of style label information, color label information, material label information, and category label information.
6. A marking processing device, characterized in that, include: The 3D model data acquisition module is used to acquire the 3D model data of the target object. The three-view determination module is used to generate three views of the target object based on the three-dimensional model data. The image compression data determination module is used to perform image preprocessing based on the three views to obtain image compression data. The label information determination module is used to input the image compressed data information into a pre-trained classification and labeling model for classification and labeling processing to obtain the label information corresponding to the target object; The three-view determination module is specifically used to perform image rendering based on the three-dimensional model data to obtain a three-dimensional model view of the target object; The front view, side view, and top view corresponding to the three-dimensional model view are defined as the three views; The image compression data determination module is specifically used to determine the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view, respectively. Image compression data is obtained by performing color channel concatenation processing based on the color channel image data of the front view, the color channel image data of the side view, and the color channel image data of the top view.
7. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the steps of the marking processing method according to any one of claims 1-5.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the marking processing method as described in any one of claims 1-5.