Marking methods, apparatus, devices, and computer-readable storage media

Abstract

This application discloses a labeling method, apparatus, device, and computer-readable storage medium. The labeling method includes: acquiring a first image to be labeled; inputting the first image to be labeled into a preset restoration model to obtain a first restored image of the first image to be labeled; determining first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image; and labeling the first image to be labeled using the first matching degree information to improve the labeling efficiency of sample data.

Description

Technical Field

[0001] This application belongs to the field of computer technology, and in particular relates to a marking method, apparatus, device, and computer-readable storage medium. Background Technology

[0002] In related technologies, perception models in the field of autonomous driving are mainly implemented using deep learning methods. This method requires a large number of labeled samples for model training, and the quality of the labeled samples affects the model performance. Therefore, it is crucial to select effective samples from massive amounts of data for labeling and incorporating them into the model for training to improve model performance and reduce manual costs. Current technologies primarily rely on manual labeling to determine sample data, which is time-consuming and inefficient. Summary of the Invention

[0003] This application provides an implementation scheme that differs from the prior art, in order to solve the technical problem of low labeling efficiency of sample data in related technologies.

[0004] Firstly, this application provides a marking method, including:

[0005] Obtain the first image to be labeled;

[0006] Input the first image to be labeled into a preset restoration model to obtain the first restored image of the first image to be labeled.

[0007] Based on the first image to be labeled and the first restored image, determine the first matching degree information between the first image to be labeled and the first restored image;

[0008] The first image to be labeled is labeled using the first matching degree information.

[0009] Secondly, this application provides a marking device, comprising:

[0010] The acquisition module is used to acquire the first image to be labeled;

[0011] The analysis module is used to input the first image to be labeled into a preset restoration model to obtain the first restored image of the first image to be labeled.

[0012] The determination module is used to determine the first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image;

[0013] The tagging module is used to tag the first image to be tagged using the first matching degree information.

[0014] Thirdly, this application provides an electronic device, including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the method described in the first aspect or any of the possible embodiments of the first aspect by executing the executable instructions.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect or any of the possible implementations of the first aspect.

[0016] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the method described in the first aspect or any of the possible implementations of the first aspect.

[0017] The labeling scheme provided in this application obtains a first image to be labeled; inputs the first image to be labeled into a preset restoration model to obtain a first restored image of the first image to be labeled; determines a first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image; and uses the first matching degree information to label the first image to be labeled. This scheme can automatically label the first image to be labeled, i.e., the sample image, thus improving the labeling efficiency of the sample image. Attached Figure Description

[0018] 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. In the drawings:

[0019] Figure 1a A schematic flowchart illustrating a marking method provided in an embodiment of this application;

[0020] Figure 1b This is a schematic diagram illustrating the working principle of a preset restoration model provided in an embodiment of this application;

[0021] Figure 1c A schematic diagram illustrating a scenario of a marking method provided in an embodiment of this application;

[0022] Figure 2 This is a schematic diagram of the structure of a marking device provided in an embodiment of this application;

[0023] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0024] The embodiments of this application are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0025] The terms "first" and "second," etc., used in the specification, claims, and drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the present application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0026] The technical solution of this application and how it solves the above-mentioned technical problems will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0027] Figure 1a The following is a flowchart illustrating a marking method provided for an exemplary embodiment of this application. The subject executing the method can be any electronic device with data processing capabilities, and the method includes at least the following steps:

[0028] S101. Obtain the first image to be labeled;

[0029] S102. Input the first image to be labeled into a preset restoration model to obtain the first restored image of the first image to be labeled.

[0030] S103. Determine the first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image;

[0031] S104. The first image to be labeled is labeled using the first matching degree information.

[0032] Specifically, the aforementioned preset restoration model can be a generative model in a GAN network model, as detailed in [reference needed]. Figure 1bAs shown, the preset restoration model can specifically include a first sub-model and a second sub-model. The first sub-model can be used to analyze the feature information in the first image to be labeled. Specifically, the feature information can be any one or more of the following: brightness distribution information, behavior information of the target object (such as a person) in the image, style information of the building in the image, and road surface material information of the road in the image.

[0033] The first sub-model can be a model with feature analysis function trained from multiple sample images, or it can be a functional module with feature analysis function trained from multiple sample images.

[0034] Accordingly, the input information of the second sub-model is the output result of the first sub-model, that is, the feature information of the first image to be labeled obtained by the first sub-model after analyzing the first image to be labeled; the second sub-model can further restore the first image to be labeled based on the feature information of the first image to be labeled to obtain the first restored image. The second sub-model can be a model with image restoration function trained on multiple sample images, or it can be a functional module with image restoration function trained on multiple sample images.

[0035] Optionally, in the aforementioned step S103, determining the first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image includes:

[0036] S01. Determine the first similarity value between the first image to be labeled and the first restored image;

[0037] S02. Use the first similarity value as the first matching degree information.

[0038] Specifically, the first similarity value between the first image to be labeled and the first restored image can be determined based on any one or more of the information in brightness, contrast, and histogram, and this application does not limit this.

[0039] Optionally, in the aforementioned step S103, determining the first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image includes:

[0040] S1. Input the first image to be labeled and the first restored image into a preset first feature analysis model to obtain the first feature information of the first image to be labeled and the second feature information of the first restored image.

[0041] S2. Determine the second similarity value between the first feature information and the second feature information;

[0042] S3. Use the second similarity value as the first matching degree information.

[0043] The first feature analysis model can be a model with feature analysis function trained from multiple sample images, specifically a neural network model.

[0044] Furthermore, the aforementioned first feature information and second feature information have the same feature type, and the feature type can be any one or more of the following: brightness distribution information, behavior information of target objects (such as people) in the image, style information of buildings in the image, and road surface material information of roads in the image.

[0045] Furthermore, the first feature information and the second feature information can be specifically text information, and the aforementioned second similarity value can refer to the text similarity value between the first feature information and the second feature information.

[0046] Optionally, inputting the first image to be labeled into a preset restoration model to obtain the first restored image of the first image to be labeled includes: inputting the first image to be labeled into the preset restoration model to obtain the first restored image of the first image to be labeled and the third feature information of the first image to be labeled. Optionally, the third feature information may be the same as the feature information in the first image to be labeled analyzed by the aforementioned first sub-model.

[0047] Furthermore, the method also includes:

[0048] The first restored image is input into a preset second feature analysis model to obtain the fourth feature information of the first restored image;

[0049] The second feature analysis model can be a model with feature analysis function trained from multiple sample images.

[0050] The fourth feature information can be text information, and the feature type of the fourth feature information can be any one or more of the following: brightness distribution information, behavior information of target objects (such as people) in the image, style information of buildings in the image, and road surface material information of roads in the image.

[0051] Further, in the aforementioned step S103, determining the first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image includes:

[0052] Determine the third similarity value between the third feature information of the first image to be labeled and the fourth feature information of the first restored image;

[0053] The third similarity value is used as the first matching degree information.

[0054] The aforementioned third feature information and fourth feature information have the same feature type, and the feature type can be any one or more of the following: brightness distribution information, behavior information of target objects (such as people) in the image, style information of buildings in the image, and road surface material information of roads in the image.

[0055] Furthermore, the third and fourth feature information can be specifically textual information, and the aforementioned fourth similarity value can refer to the textual similarity value between the third and fourth feature information.

[0056] Here, the first matching degree information is the latent space features corresponding to the first image to be labeled and the first restored image determined by the first sub-model. It focuses more on the differences in the actual content of the images and is not sensitive to small changes in the images. Therefore, it can solve the problem that the first sub-model is susceptible to noise and has better robustness.

[0057] Optionally, in the aforementioned step S104, labeling the first image to be labeled using the first matching degree information includes:

[0058] If the target similarity value indicated by the first matching degree information is greater than the preset similarity value, then the first image to be labeled is labeled as an image that meets the requirements;

[0059] If the target similarity value indicated by the first matching information is not greater than the preset similarity value, then the first image to be labeled is marked as an image that does not meet the requirements.

[0060] Specifically, the target similarity value indicated by the first matching degree information is the similarity value used to determine the first matching degree information. For example, if the first matching degree information is determined by the first similarity value, then the target similarity value is the first similarity value; if the first matching degree information is determined by the second similarity value, then the target similarity value is the second similarity value.

[0061] Furthermore, when marking the first image to be marked, a specific correspondence between the first image to be marked and its marking identifier can be created. For example, marking the first image to be marked as a compliant image includes associating the first image to be marked with a preset first identifier corresponding to compliance; marking the second image to be marked as a non-compliant image includes associating the second image to be marked with a preset second identifier corresponding to non-compliance.

[0062] Furthermore, the first image to be labeled is an image from the image set to be labeled, and the method further includes:

[0063] Detect whether all images in the set of images to be labeled have been labeled; if not, obtain the second image to be labeled.

[0064] The second image to be labeled is input into the preset restoration model to obtain the second restored image of the second image to be labeled.

[0065] Based on the second image to be labeled and the second restored image, determine the second matching degree information between the second image to be labeled and the second image to be restored;

[0066] The second image to be labeled is labeled using the second matching degree information.

[0067] The image set to be labeled is a collection of multiple images to be labeled. The second image to be labeled is any image in the image set that has not been labeled. The method for labeling the second image is the same as the method for labeling the first image, and will not be described again here.

[0068] Furthermore, the method also includes:

[0069] If it is detected that all images in the image set to be labeled have been labeled;

[0070] The set of images in the image set to be labeled as not meeting the requirements is then taken as the target image set;

[0071] The target image set and the labeling results corresponding to each image in the target image set are sent to the target device.

[0072] Specifically, the target image set and the labeling results corresponding to each image in the target image set are sent to the target device, so that the administrators of the target device can view the target image set and the labeling results corresponding to each image in the target image set.

[0073] It should be noted that the feature analysis modules involved in this application may be the same or different, and this application does not limit them.

[0074] This application enables automatic labeling of a first image to be labeled, and the first image to be labeled as not meeting the requirements can be used as a new sample image.

[0075] Furthermore, the first matching degree information may also be a weighted result of the aforementioned first similarity value, second similarity value, and third similarity value, and this application does not limit this.

[0076] Furthermore, when the solution of this application is used for the analysis of driving scenario data, after obtaining the first image to be labeled, it also includes a process for determining a preset restoration model. The above method also includes:

[0077] Obtain the image type information input by the user;

[0078] Based on the image type information, a preset restoration model corresponding to the image type information is selected from multiple candidate restoration models.

[0079] The aforementioned image type information may include any one or more of the following: brightness distribution (or weather), behavior analysis, architectural style, and road surface material. The brightness distribution, behavior analysis, architectural style, and road surface material correspond to the brightness distribution information, the behavior information of the target object (such as a person) in the image, the style information of the building in the image, and the road surface material information of the road in the image, respectively, in the aforementioned first feature information.

[0080] This approach can further improve the efficiency of image labeling.

[0081] Furthermore, the determination of image type information can be based on user instructions or on a relevant neural network model; this application does not limit this.

[0082] It should be noted that the images that the preset restoration model cannot successfully restore in this application (i.e., the first unlabeled images marked as not meeting the requirements) are the selected sample data. This selected sample data is used to train a new model or as supplementary sample data for the preset restoration model to further improve the function of the preset restoration model so that it can restore more types of images.

[0083] This patent provides an automatic sample labeling method based on OOD (out-of-distribution) detection. This method is simple, efficient, and improves sample data filtering and model performance.

[0084] This application is primarily used in target detection algorithms for autonomous driving perception systems. When the number of known categories is large, the aforementioned preset restoration model can be controlled to fully learn the distribution characteristics of normal data (corresponding to images that meet the requirements) to identify abnormal images (i.e., images that do not meet the requirements). This scheme can select missing sample data from a new unlabeled dataset, label them, and train a new model. The aim is to detect samples with distributions different from those already added to the training dataset. These selected samples are then added to the model training to improve its robustness.

[0085] The labeling scheme provided in this application obtains a first image to be labeled; inputs the first image to be labeled into a preset restoration model to obtain a first restored image of the first image to be labeled; determines a first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image; and uses the first matching degree information to label the first image to be labeled. This scheme can automatically label the first image to be labeled, i.e., the sample image, thus improving the labeling efficiency of the sample image.

[0086] The following section provides further explanation of this solution in specific scenarios:

[0087] Scene 1

[0088] See Figure 1c As shown, a network structure of encoder 1-decoder-encoder 2 can be specifically adopted, simultaneously learning two mapping relationships: "original image (corresponding to the aforementioned first image to be analyzed) -> reconstructed image (corresponding to the aforementioned first restored image)" and "encoding of the original image -> encoding of the reconstructed image". This method not only constrains the appearance of the generated image (image -> image) but also constrains the image content (image encoding -> image encoding).

[0089] Optionally, for Figure 1c During the training phase, the data distribution of normal data can be learned only to obtain a target model that includes a preset restoration model, a first feature analysis model, and a second feature analysis model (i.e., encoder 2).

[0090] During the testing phase, normal and abnormal data can be used to determine whether the restored data is the same as or close to the original data based on the target model. If so, it indicates that the data being tested is normal data; otherwise, it is abnormal data.

[0091] Scene 2

[0092] Training set: The experiment selected 900 images each from nighttime, sunny, and cloudy days, totaling 2700 images, as normal samples (i.e., sample images that meet the requirements);

[0093] Test set: 300 images were selected as normal samples for the test set, including 100 images each from nighttime, sunny days, and cloudy days. 300 images were selected as abnormal samples (i.e., images that meet the requirements) from underground days, rainy days, and snowy days.

[0094] Experimental results: Using AUC (Area Under Curve) as the binary classification evaluation metric, the test results show an accuracy of 0.78 when the image resolution is 64*64 and 0.8 when the image resolution is 128*128.

[0095] Scene 3

[0096] Training set: 1100 image samples were collected from the first region;

[0097] Test set: 200 image samples from the first region and 200 image samples from the second region;

[0098] Experimental results:

[0099] The image resolution is 128*128, and the test results show that the accuracy is 0.96.

[0100] Experimental conclusion:

[0101] This solution can effectively achieve automatic sample selection, reducing the time cost of manual random sample selection.

[0102] Figure 2 A schematic diagram of a marking device provided for an exemplary embodiment of this application; wherein the device includes: an acquisition module 21, an analysis module 22, a determination module 23, and a marking module 24; wherein:

[0103] Acquisition module 21 is used to acquire the first image to be labeled;

[0104] Analysis module 22 is used to input the first image to be labeled into a preset restoration model to obtain the first restored image of the first image to be labeled;

[0105] The determining module 23 is used to determine the first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image;

[0106] The labeling module 24 is used to label the first image to be labeled using the first matching degree information.

[0107] Optionally, when the above-described apparatus is used to determine the first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image, it is specifically used for:

[0108] Determine a first similarity value between the first image to be labeled and the first restored image;

[0109] The first similarity value is used as the first matching degree information.

[0110] Optionally, when the above-described apparatus is used to determine the first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image, it is specifically used for:

[0111] The first image to be labeled and the first restored image are respectively input into a preset first feature analysis model to obtain the first feature information of the first image to be labeled and the second feature information of the first restored image.

[0112] Determine a second similarity value between the first feature information and the second feature information;

[0113] The second similarity value is used as the first matching degree information.

[0114] Optionally, when the above-mentioned device is used to input the first image to be labeled into a preset restoration model to obtain a first restored image of the first image to be labeled, it is specifically used to: input the first image to be labeled into the preset restoration model to obtain the first restored image of the first image to be labeled and the third feature information of the first image to be labeled; further, the above-mentioned device is also used to:

[0115] The first restored image is input into a preset second feature analysis model to obtain the fourth feature information of the first restored image;

[0116] Furthermore, when the above-mentioned device is used to determine the first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image, it is specifically used for:

[0117] Determine the third similarity value between the third feature information of the first image to be labeled and the fourth feature information of the first restored image;

[0118] The third similarity value is used as the first matching degree information.

[0119] Optionally, when the above-described device is used to label the first image to be labeled using the first matching degree information, it is specifically used for:

[0120] If the target similarity value indicated by the first matching degree information is greater than the preset similarity value, then the first image to be labeled is labeled as an image that meets the requirements;

[0121] If the target similarity value indicated by the first matching information is not greater than the preset similarity value, then the first image to be labeled is marked as an image that does not meet the requirements.

[0122] Optionally, the first image to be labeled is an image from a set of images to be labeled, and the apparatus is further configured to:

[0123] Detect whether all images in the set of images to be labeled have been labeled; if not, obtain the second image to be labeled.

[0124] The second image to be labeled is input into the preset restoration model to obtain the second restored image of the second image to be labeled.

[0125] Based on the second image to be labeled and the second restored image, determine the second matching degree information between the second image to be labeled and the second image to be restored;

[0126] The second image to be labeled is labeled using the second matching degree information.

[0127] Optionally, the above-mentioned device is also used for:

[0128] If it is detected that all images in the image set to be labeled have been labeled;

[0129] The set of images in the image set to be labeled as not meeting the requirements is then taken as the target image set;

[0130] The target image set and the labeling results corresponding to each image in the target image set are sent to the target device.

[0131] It should be understood that the device embodiments and method embodiments can correspond to each other, and similar descriptions can be referred to the method embodiments. To avoid repetition, they will not be repeated here. Specifically, the device can execute the above method embodiments, and the foregoing and other operations and / or functions of each module in the device correspond to the corresponding processes in the various methods in the above method embodiments, which will not be repeated here for the sake of brevity.

[0132] The apparatus of this application embodiment has been described above from the perspective of functional modules in conjunction with the accompanying drawings. It should be understood that this functional module can be implemented in hardware, in software instructions, or in a combination of hardware and software modules. Specifically, the steps of the method embodiments in this application can be completed by integrated logic circuits in the processor's hardware and / or by software instructions. The steps of the method disclosed in this application embodiment can be directly embodied as being executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. Optionally, the software module can reside in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps in the above method embodiments.

[0133] Figure 3 This is a schematic block diagram of an electronic device provided in an embodiment of this application. The electronic device may include:

[0134] The system includes a memory 301 and a processor 302. The memory 301 stores computer programs and transfers the program code to the processor 302. In other words, the processor 302 can retrieve and run the computer programs from the memory 301 to implement the methods described in the embodiments of this application.

[0135] For example, the processor 302 can be used to execute the above-described method embodiments according to instructions in the computer program.

[0136] In some embodiments of this application, the processor 302 may include, but is not limited to:

[0137] General-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0138] In some embodiments of this application, the memory 301 includes, but is not limited to:

[0139] Volatile memory and / or non-volatile memory. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), and Direct Rambus RAM (DR RAM).

[0140] In some embodiments of this application, the computer program may be divided into one or more modules, which are stored in the memory 301 and executed by the processor 302 to perform the method provided in this application. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0141] like Figure 3 As shown, the electronic device may further include:

[0142] Transceiver 303, which can be connected to processor 302 or memory 301.

[0143] The processor 302 can control the transceiver 303 to communicate with other devices; specifically, it can send information or data to other devices or receive information or data sent by other devices. The transceiver 303 may include a transmitter and a receiver. The transceiver 303 may further include antennas, and the number of antennas may be one or more.

[0144] It should be understood that the various components in the electronic device are connected through a bus system, which includes a data bus, a power bus, a control bus, and a status signal bus.

[0145] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a computer, enables the computer to perform the methods of the above-described method embodiments. Alternatively, embodiments of this application also provide a computer program product containing instructions that, when executed by a computer, cause the computer to perform the methods of the above-described method embodiments.

[0146] When implemented using software, it can be implemented entirely or partially as a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital video disc (DVD)), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0147] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0148] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0149] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; 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. For example, the functional modules in the various embodiments of this application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0150] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A marking method, characterized in that, include: Obtain the first image to be labeled; Input the first image to be labeled into a preset restoration model to obtain the first restored image of the first image to be labeled. Based on the first image to be labeled and the first restored image, determine the first matching degree information between the first image to be labeled and the first restored image; The first image to be labeled is labeled using the first matching degree information; Determining the first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image includes: The first image to be labeled and the first restored image are respectively input into a preset first feature analysis model to obtain the first feature information of the first image to be labeled and the second feature information of the first restored image. Determine a second similarity value between the first feature information and the second feature information; The second similarity value is used as the first matching degree information; The first feature information and the second feature information have the same feature type, and the feature type is any one or more of the following: brightness distribution information, behavior information of target objects in the image, style information of buildings in the image, and road surface material information of roads in the image; the first feature information and the second feature information are text information, and the second similarity value is the text similarity value between the first feature information and the second feature information. or The method of inputting the first image to be labeled into a preset restoration model to obtain a first restored image of the first image to be labeled includes: inputting the first image to be labeled into the preset restoration model to obtain the first restored image of the first image to be labeled and the third feature information of the first image to be labeled, the method further includes: The first restored image is input into a preset second feature analysis model to obtain the fourth feature information of the first restored image; Determining the first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image includes: Determine the third similarity value between the third feature information of the first image to be labeled and the fourth feature information of the first restored image; The third similarity value is used as the first matching degree information; The third feature information and the fourth feature information have the same feature type, and the feature type is any one or more of the following: brightness distribution information, behavior information of target objects in the image, style information of buildings in the image, and road surface material information of roads in the image; the third feature information and the fourth feature information are text information, and the third similarity value is the text similarity value between the third feature information and the fourth feature information.

2. The method according to claim 1, characterized in that, Labeling the first image to be labeled using the first matching degree information includes: If the target similarity value indicated by the first matching degree information is greater than the preset similarity value, then the first image to be labeled is labeled as an image that meets the requirements; If the target similarity value indicated by the first matching information is not greater than the preset similarity value, then the first image to be labeled is marked as an image that does not meet the requirements.

3. The method according to claim 2, characterized in that, The first image to be labeled is an image from the image set to be labeled, and the method further includes: Detect whether all images in the set of images to be labeled have been labeled; if not, obtain the second image to be labeled. The second image to be labeled is input into the preset restoration model to obtain the second restored image of the second image to be labeled. Based on the second image to be labeled and the second restored image, determine the second matching degree information between the second image to be labeled and the second restored image; The second image to be labeled is labeled using the second matching degree information.

4. The method according to claim 3, characterized in that, The method further includes: If it is detected that all images in the image set to be labeled have been labeled; The set of images in the image set to be labeled as not meeting the requirements is then taken as the target image set; The target image set and the labeling results corresponding to each image in the target image set are sent to the target device.

5. A marking device, characterized in that, include: The acquisition module is used to acquire the first image to be labeled; The analysis module is used to input the first image to be labeled into a preset restoration model to obtain the first restored image of the first image to be labeled. The determination module is used to determine the first matching degree information between the first image to be labeled and the first restored image based on the first image to be labeled and the first restored image; A labeling module is used to label the first image to be labeled using the first matching degree information; The determining module is further configured to input the first image to be labeled and the first restored image into a preset first feature analysis model respectively, and obtain the first feature information of the first image to be labeled and the second feature information of the first restored image respectively; Determine a second similarity value between the first feature information and the second feature information; use the second similarity value as the first matching degree information; Wherein, the first feature information and the second feature information have the same feature type and the feature type is any one or more of the following: brightness distribution information, behavior information of target objects in the image, style information of buildings in the image, and pavement material information of roads in the image; the first feature information and the second feature information are text information, and the second similarity value is the text similarity value between the first feature information and the second feature information. or The analysis module is further configured to input the first image to be labeled into a preset restoration model to obtain the first restored image of the first image to be labeled and the third feature information of the first image to be labeled, and input the first restored image into a preset second feature analysis model to obtain the fourth feature information of the first restored image. The determining module is further configured to determine a third similarity value between the third feature information of the first image to be labeled and the fourth feature information of the first restored image; and use the third similarity value as the first matching degree information; Wherein, the third feature information and the fourth feature information have the same feature type, and the feature type is any one or more of the following: brightness distribution information, behavior information of target objects in the image, style information of buildings in the image, and pavement material information of roads in the image; the third feature information and the fourth feature information are text information, and the third similarity value is the text similarity value between the third feature information and the fourth feature information.

6. An electronic device, characterized in that, include: processor; as well as Memory for storing the executable instructions of the processor; The processor is configured to perform the method of any one of claims 1-4 by executing the executable instructions.

7. 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 method of any one of claims 1-4.

Images