Vehicle damage level evaluation method and device, computer equipment and storage medium

Abstract

The invention relates to the field of artificial intelligence, and provides a vehicle damage level evaluation method, device and equipment and a computer storage medium, and the method comprises the steps: obtaining a to-be-evaluated damaged vehicle image, and inputting the damaged vehicle image into a preset target detection model, the target detection model comprises an identification network used for identifying a target object and a classification head used for classifying the target object; acquiring at least one target object obtained by identifying the damaged vehicle image by the identification network; respectively inputting the target object into the classification head and a preset regression head, and obtaining a first evaluation result output by the classification head and a second evaluation result output by the regression head; and performing result fusion on the first evaluation result and the second evaluation result to obtain a target evaluation result of the vehicle damage level.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to a method, device, computer equipment and storage medium for assessing the level of damage to a vehicle. Background Art

[0002] Vehicle damage assessment, that is, the process of assessing the extent of damage to a vehicle in an accident and the cost of repairing it, is an important part of insurance claims and vehicle repair. Existing object detection models can well identify damage in damaged vehicle images, but there may be a class imbalance problem when classifying the degree of damage, resulting in a high misclassification rate for certain damages. There is an urgent need for a method that can improve the accuracy of vehicle damage assessment. Summary of the invention

[0003] The main purpose of the present application is to provide a vehicle damage level assessment method, device, equipment and computer storage medium, aiming to improve the accuracy of vehicle damage level assessment.

[0004] In a first aspect, the present application provides a method for assessing a vehicle damage level, the method comprising the following steps:

[0005] Acquire an image of a damaged vehicle to be assessed, and input the damaged vehicle image into a preset target detection model, wherein the target detection model includes a recognition network for identifying a target object and a classification head for classifying the target object;

[0006] Acquire at least one target object obtained by the recognition network identifying the damaged vehicle image;

[0007] Input the target object into the classification head and the preset regression head respectively, and obtain a first evaluation result output by the classification head and a second evaluation result output by the regression head;

[0008] The first assessment result and the second assessment result are fused to obtain a target assessment result of the vehicle damage level.

[0009] In a second aspect, the present application further provides a vehicle damage level assessment device, the vehicle damage level assessment device comprising:

[0010] An image acquisition module, used for acquiring an image of a damaged vehicle to be assessed, and inputting the damaged vehicle image into a preset target detection model, wherein the target detection model includes a recognition network for identifying a target object and a classification head for classifying the target object;

[0011] A target object determination module, used to obtain at least one target object obtained by the recognition network identifying the damaged vehicle image;

[0012] An evaluation result determination module, used to input the target object into the classification head and the preset regression head respectively, and obtain a first evaluation result output by the classification head and a second evaluation result output by the regression head;

[0013] The result fusion module is used to fuse the first assessment result and the second assessment result to obtain a target assessment result of the vehicle damage level.

[0014] In a third aspect, the present application also provides a computer device, comprising a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, the vehicle damage level assessment method as described above is implemented.

[0015] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the vehicle damage level assessment method as described above is implemented.

[0016] The present application provides a method, device, equipment and computer storage medium for assessing the level of damage to a vehicle. The present application obtains an image of a damaged vehicle to be assessed, and inputs the damaged vehicle image into a preset target detection model, wherein the target detection model includes a recognition network for identifying a target object and a classification head for classifying the target object; obtains at least one target object obtained by the recognition network identifying the damaged vehicle image; inputs the target object into the classification head and a preset regression head respectively, and obtains a first assessment result output by the classification head and a second assessment result output by the regression head; and fuses the first assessment result and the second assessment result to obtain a target assessment result of the level of damage to the vehicle. Since different assessment methods are combined to determine the target assessment result of the level of damage to the vehicle, the accuracy of the assessment of the level of damage to the vehicle is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 A schematic diagram of a flow chart of a method for assessing a vehicle damage level provided in one embodiment of the present application;

[0019] Figure 2A diagram showing a usage scenario of a method for assessing a vehicle damage level provided in one embodiment of the present application;

[0020] Figure 3 A schematic block diagram of a vehicle damage level assessment device provided in one embodiment of the present application;

[0021] Figure 4 The present invention is a block diagram showing the structure of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0023] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0024] Embodiments of the present application provide a method, apparatus, computer device, and computer-readable storage medium for assessing a vehicle damage level.

[0025] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0026] Please refer to Figure 1 , Figure 1 A flow chart of a method for assessing the level of vehicle damage provided in an embodiment of the present application. The method for assessing the level of vehicle damage can be used in a terminal or a server to determine the level of vehicle damage of a damaged vehicle based on an image of the damaged vehicle. The terminal can be an electronic device such as a mobile phone, a tablet computer, a laptop computer, a desktop computer, a personal digital assistant, and a wearable device; the server can be an independent server, a server cluster, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0027] Please refer to Figure 2 , Figure 2 This is a usage scenario diagram provided by an embodiment of the present application. Figure 2 As shown, the damage level assessment model provided in the embodiment of the present application combines the prediction results of the regression head with the target detection model to predict the target assessment results of the vehicle damage level, thereby realizing automatic damage assessment based on the damaged vehicle image and improving the efficiency and accuracy of vehicle damage assessment.

[0028] like Figure 1 As shown, the vehicle damage level assessment method includes steps S101 to S104.

[0029] Step S101: Acquire an image of a damaged vehicle to be assessed, and input the damaged vehicle image into a preset target detection model, wherein the target detection model includes a recognition network for identifying a target object and a classification head for classifying the target object.

[0030] For example, vehicle damage assessment is the basis for handling automobile insurance claims in the insurance business. Evaluating the repair costs of damaged vehicles based on accurate damage assessment results helps insurance companies reasonably calculate the amount of compensation and ensure the rational use of insurance funds. In actual operations, manual damage assessment relies on the experience and skills of professional damage assessors, and the damage assessment results can be flexibly adjusted according to specific circumstances. However, the efficiency of damage assessment may be limited by the location of the accident and the personal experience of the damage assessor, which may lead to high operating costs and easy disputes. The automatic damage assessment technology based on artificial intelligence can significantly improve the efficiency of claims processing, reduce errors caused by human factors, and improve customer experience.

[0031] For example, in order to realize automatic damage assessment, it is necessary to first obtain the image of the damaged vehicle to be assessed, so as to input the damaged vehicle image into the preset model and determine the target assessment result of the vehicle damage level according to the damaged vehicle image. The damaged vehicle image includes the whole vehicle image and the close-up image of the damaged part from multiple angles, ensuring the multi-angle view of the whole vehicle and the damaged part to provide more comprehensive information.

[0032] For example, an image quality detection may be performed when acquiring the damaged vehicle image, and the user is allowed to upload the damaged vehicle image only when the image quality of the damaged vehicle image meets a preset image quality condition.

[0033] Exemplarily, the type of target detection model is not limited herein. For example, a real-time target detection model may be used, such as various versions of the YOLO model. The YOLO model has a high accuracy in target detection tasks and reduces computational effort and memory usage while maintaining high accuracy, and can accurately and efficiently identify the type and location of vehicle damage.

[0034] Step S102: Acquire at least one target object obtained by the recognition network through recognition of the damaged vehicle image.

[0035] Exemplarily, the target detection model in step S101 includes a recognition network for identifying the target object and a classification head for classifying the target object. Specifically, the recognition network is used to identify the damaged parts in the damaged vehicle image, and the classification head can classify the damaged parts to determine the damage level of the damaged parts.

[0036] Exemplarily, taking the YOLO model as an example, the recognition network can be the backbone network in the YOLO model. Specifically, YOLOv1 uses a custom network structure similar to GoogLeNet as its backbone network; YOLOv2 (ie YOLO9000) uses an improved version of Darknet-19; YOLOv3 uses a deeper Darknet-53; and each backbone network is also set in the updated version of the YOLO model.

[0037] Exemplarily, the recognition network in the target detection model can extract image features and recognize the target object based on the image features. In an embodiment of the present application, the recognition network is used to recognize the damaged part in the damaged vehicle image.

[0038] Step S103: input the target object into the classification head and the preset regression head respectively, and obtain a first evaluation result output by the classification head and a second evaluation result output by the regression head.

[0039] Exemplarily, after the target object is identified, the type of the target object is classified by a classification head, wherein the number of categories of the classification head can be determined according to actual needs, thereby determining whether the classification head performs a binary classification task or a multi-class classification task. For example, according to the degree of repair of the repair plan, the vehicle damage level is divided into "painting, small repair, medium repair, large repair, replacement" from light to heavy, and represented by category codes "1, 2, 3, 4, 5" respectively, for example, 0 represents painting, and so on, which will not be repeated here.

[0040] Exemplarily, in order to improve the accuracy of vehicle damage level assessment, increase the diversity of the model so as to capture different features of the data, and help reduce the risk of decision-making errors, after the recognition network identifies the target object, in addition to inputting the target object into the classification head to obtain a first assessment result, the target object is also input into the regression head to obtain a second assessment result, so as to fuse the first assessment result and the second assessment result to obtain a more accurate target assessment result.

[0041] In some implementations, inputting the target object into the classification head and the preset regression head respectively, and obtaining a first evaluation result output by the classification head and a second evaluation result output by the regression head, comprises:

[0042] Obtaining the confidence of each sub-image corresponding to the damaged vehicle image;

[0043] When the confidence is greater than a preset confidence threshold, the sub-image is determined as the target object and input into the classification head and the preset regression head.

[0044] Exemplarily, when performing target object recognition, the recognition network will segment the damaged vehicle image into multiple sub-images of preset sizes, determine the probability of containing the target object in the sub-image, and output it as the confidence level of each target object.

[0045] Exemplarily, when the confidence level is greater than a preset confidence level threshold, the sub-image is taken as a target object and input into a preset classification head and a regression head respectively.

[0046] In some implementations, inputting the target object into the classification head and the preset regression head respectively, and obtaining a first evaluation result output by the classification head and a second evaluation result output by the regression head, comprises:

[0047] The regression prediction result of the regression head is obtained, and the second evaluation result Y=sigmoid(x)*n*α+β is determined according to the following formula, wherein Y is the second evaluation result, x is the regression prediction result, n is the number of vehicle damage levels, α is the first hyperparameter, and β is the second hyperparameter.

[0048] Exemplarily, the output of the classification head is a discrete category label, each label represents a category, and the output of the regression head is a continuous value, which can be any real number. In order to make the output of the regression head consistent with the preset vehicle damage level, the regression prediction result output by the regression head is mapped to the (0,1) interval through the Sigmoid function, and then multiplied by the number of categories n, and then adjusted through two hyperparameters α and β. The output range of the result. In the vehicle damage level assessment method provided in the embodiment of the present application, n can be, for example, 5, of course, it is not limited to this, and is not limited here.

[0049] Exemplarily, the result obtained according to the above formula is a real number between 0 and 5. Since the vehicle damage level has a discrete characteristic, the calculation result can be rounded off and used as the second assessment result.

[0050] Step S104: Fusing the first assessment result and the second assessment result to obtain a target assessment result of the vehicle damage level.

[0051] Exemplarily, since different prediction methods may perform better in different aspects, after respectively obtaining the first evaluation result of the classification head and the second evaluation result of the regression head in step S103, the first evaluation result of the classification head and the second evaluation result of the regression head are integrated to reduce the overall error of the model, and the first evaluation result and the second evaluation result can be fused to obtain the target evaluation result.

[0052] In some implementations, the fusing the first assessment result and the second assessment result to obtain the target assessment result of the vehicle damage level includes:

[0053] Get the preset result fusion strategy;

[0054] Based on the result fusion strategy, the target assessment result is determined according to the first assessment result and / or the second assessment result.

[0055] For example, a result fusion strategy can be set according to actual needs, such as determining the larger or smaller of the first assessment result and the second assessment result as the target assessment result, or determining the target assessment result by combining the first assessment result and the second assessment result, thereby improving the accuracy of the target assessment result.

[0056] In some embodiments, determining the target assessment result based on the result fusion strategy according to the first assessment result and / or the second assessment result includes:

[0057] In a case where the level difference between the first assessment result and the second assessment result is less than a preset threshold, based on the result fusion strategy, determining the first assessment result or the second assessment result as the target assessment result;

[0058] In a case where the level difference between the first assessment result and the second assessment result is greater than or equal to a preset threshold, a median level of the first assessment result and the second assessment result is determined as the target assessment result.

[0059] Exemplarily, if the gap between the first assessment result and the second assessment result is small, selecting one of the first assessment result and the second assessment result as the target assessment result will not produce significant difference. Based on the result fusion strategy preset by the user, such as a strategy of taking a larger value or a smaller value, a larger level or a smaller level can be selected from the first assessment result and the second assessment result as the target assessment result.

[0060] For example, if there is a large gap between the first assessment result and the second assessment result, it indicates that both may have certain errors, and using the first assessment result or the second assessment result as the target assessment result will lead to a large difference. At this time, the first assessment result and the second assessment result can be combined, and the median level of the two can be used as the target assessment result of the vehicle damage level.

[0061] For example, the preset threshold of the level difference can be 2. If the first assessment result and the second assessment result differ by one level, for example, the first assessment result is 1 and the second assessment result is 2, then the target assessment result is determined from the two; conversely, if the first assessment result and the second assessment result differ by more than two levels, for example, the first assessment result is 1 and the second assessment result is 3, then 2 between the two is confirmed as the target assessment result.

[0062] In some embodiments, the method further comprises:

[0063] Obtaining a first evaluation result and a second evaluation result of a preset sample image, and determining first loss information of the classification head and second loss information of the regression head according to the annotation information corresponding to the sample image and the first evaluation result and the second evaluation result;

[0064] determining target loss information according to the first loss information and the second loss information;

[0065] When the target loss information does not meet the preset conditions, the classification head is adjusted according to the first loss information, and the first hyperparameter and the second hyperparameter of the regression head are adjusted according to the second loss information until the target loss information meets the preset conditions.

[0066] Exemplarily, when training the damage level assessment model, the loss value of the classification head and the loss value of the regression head are obtained respectively, and the target loss information is determined based on the combination of the two, thereby reducing the overall error of the damage level assessment model.

[0067] Exemplarily, a sample image pre-labeled with a vehicle damage level as annotation information is obtained, the sample image is input into a damage level assessment model and a first assessment result and a second assessment result of the sample image are obtained, first loss information of a classification head is determined according to the first assessment result and the annotation information, second loss information of the classification head is determined according to the second assessment result and the annotation information, and target loss information of the damage level assessment model is calculated according to the first loss information and the second loss information. When the target loss information does not meet preset conditions, the parameters of the classification head and the regression head are adjusted respectively to reduce the first loss information and the second loss information.

[0068] The adjustment of the parameters of the regression head also includes the adjustment of the first hyperparameter α and the second hyperparameter β. At the beginning of training, the value of the first hyperparameter α is preset to 1, and the value of the second hyperparameter β is preset to 0, and the values ​​of the two are adjusted during the training process so that the regression head outputs reasonable results.

[0069] In some implementations, determining target loss information according to the first loss information and the second loss information includes:

[0070] Determine first loss data according to the product of the first loss information and a first preset coefficient, and determine second loss data according to the product of the second loss information and a second preset coefficient;

[0071] The target loss information is determined according to the sum of the first loss data and the second loss data.

[0072] Exemplarily, the first loss information and the second information are weighted and summed to obtain the target loss information. Specifically, the target loss information Loss=θ1L det +θ2L reg , where θ1 is the first preset coefficient, L det is the first loss information of the classification head, θ2 is the second preset coefficient, L reg is the second loss information of the regression head.

[0073] Exemplarily, the calculation method of the first loss information and the second loss information is not limited here, for example, it can be determined based on the mean square error MSE or the absolute error L1, of course, it is not limited to this and is not limited here.

[0074] The vehicle damage level assessment method provided in the above embodiment obtains a damaged vehicle image to be assessed, and inputs the damaged vehicle image into a preset target detection model, wherein the target detection model includes a recognition network for identifying target objects and a classification head for classifying the target objects; obtains at least one target object obtained by the recognition network identifying the damaged vehicle image; inputs the target object into the classification head and the preset regression head respectively, and obtains a first assessment result output by the classification head and a second assessment result output by the regression head; and fuses the first assessment result and the second assessment result to obtain a target assessment result of the vehicle damage level. The target assessment result of the vehicle damage level can be determined by combining different assessment methods, thereby improving the accuracy of the vehicle damage level assessment.

[0075] See also Figure 3 , Figure 31 is a schematic diagram of a vehicle damage level assessment device provided in an embodiment of the present application. The vehicle damage level assessment device can be configured in a server or a terminal to execute the aforementioned vehicle damage level assessment method.

[0076] like Figure 3 As shown, the vehicle damage level assessment device includes: an image acquisition module 110 , a target object determination module 120 , an assessment result determination module 130 , and a result fusion module 140 .

[0077] An image acquisition module 110 is used to acquire an image of a damaged vehicle to be assessed, and input the damaged vehicle image into a preset target detection model, wherein the target detection model includes a recognition network for identifying a target object and a classification head for classifying the target object;

[0078] A target object determination module 120, configured to obtain at least one target object identified by the recognition network on the damaged vehicle image;

[0079] An evaluation result determination module 130, used to input the target object into the classification head and the preset regression head respectively, and obtain a first evaluation result output by the classification head and a second evaluation result output by the regression head;

[0080] The result fusion module 140 is used to fuse the first assessment result and the second assessment result to obtain a target assessment result of the vehicle damage level.

[0081] In some implementations, the result fusion module 140, in the process of fusing the first assessment result and the second assessment result to obtain the target assessment result of the vehicle damage level, is used to implement:

[0082] Get the preset result fusion strategy;

[0083] Based on the result fusion strategy, the target assessment result is determined according to the first assessment result and / or the second assessment result.

[0084] In some implementations, the result fusion module 140, in the process of determining the target assessment result according to the first assessment result and / or the second assessment result based on the result fusion strategy, is used to implement:

[0085] In a case where the level difference between the first assessment result and the second assessment result is less than a preset threshold, based on the result fusion strategy, determining the first assessment result or the second assessment result as the target assessment result;

[0086] In a case where the level difference between the first assessment result and the second assessment result is greater than or equal to a preset threshold, a median level of the first assessment result and the second assessment result is determined as the target assessment result.

[0087] In some implementations, the evaluation result determination module 130, in the process of inputting the target object into the classification head and the preset regression head respectively, and obtaining the first evaluation result output by the classification head and the second evaluation result output by the regression head, is used to implement:

[0088] Obtaining the confidence of each sub-image corresponding to the damaged vehicle image;

[0089] When the confidence is greater than a preset confidence threshold, the sub-image is determined as the target object and input into the classification head and the preset regression head.

[0090] In some implementations, the evaluation result determination module 130, in the process of inputting the target object into the classification head and the preset regression head respectively, and obtaining the first evaluation result output by the classification head and the second evaluation result output by the regression head, is used to implement:

[0091] The regression prediction result of the regression head is obtained, and the second evaluation result Y=sigmoid(x)*n*α+β is determined according to the following formula, wherein Y is the second evaluation result, x is the regression prediction result, n is the number of vehicle damage levels, α is the first hyperparameter, and β is the second hyperparameter.

[0092] In some embodiments, the vehicle damage level assessment device, in the process of implementing the vehicle damage level assessment method, is used to implement:

[0093] Obtaining a first evaluation result and a second evaluation result of a preset sample image, and determining first loss information of the classification head and second loss information of the regression head according to the annotation information corresponding to the sample image and the first evaluation result and the second evaluation result;

[0094] determining target loss information according to the first loss information and the second loss information;

[0095] When the target loss information does not meet the preset conditions, the classification head is adjusted according to the first loss information, and the first hyperparameter and the second hyperparameter of the regression head are adjusted according to the second loss information until the target loss information meets the preset conditions.

[0096] In some implementations, the vehicle damage level assessment device, in the process of determining the target loss information according to the first loss information and the second loss information, is used to implement:

[0097] Determine first loss data according to the product of the first loss information and a first preset coefficient, and determine second loss data according to the product of the second loss information and a second preset coefficient;

[0098] The target loss information is determined according to the sum of the first loss data and the second loss data.

[0099] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and modules and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0100] The method and apparatus of the present application can be used in many general or special computing system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0101] Exemplarily, the above method and apparatus may be implemented in the form of a computer program. The computer program may be implemented in Figure 4 Runs on the computer device shown.

[0102] See also Figure 4 , Figure 4 A schematic block diagram of the structure of a computer device provided in an embodiment of the present application. The computer device may be a server or a terminal.

[0103] like Figure 4 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus, wherein the memory may include a storage medium and an internal memory.

[0104] The storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any one of the vehicle damage level assessment methods.

[0105] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0106] The internal memory provides an environment for the operation of the computer program in the storage medium. When the computer program is executed by the processor, the processor can execute any method for assessing the vehicle damage level.

[0107] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will appreciate that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0108] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0109] In one embodiment, the processor is used to run a computer program stored in the memory to implement the following steps:

[0110] Acquire an image of a damaged vehicle to be assessed, and input the damaged vehicle image into a preset target detection model, wherein the target detection model includes a recognition network for identifying a target object and a classification head for classifying the target object;

[0111] Acquire at least one target object obtained by the recognition network identifying the damaged vehicle image;

[0112] Input the target object into the classification head and the preset regression head respectively, and obtain a first evaluation result output by the classification head and a second evaluation result output by the regression head;

[0113] The first assessment result and the second assessment result are fused to obtain a target assessment result of the vehicle damage level.

[0114] In one embodiment, when the processor implements the fusion of the first assessment result and the second assessment result to obtain the target assessment result of the vehicle damage level, it is used to implement:

[0115] Get the preset result fusion strategy;

[0116] Based on the result fusion strategy, the target assessment result is determined according to the first assessment result and / or the second assessment result.

[0117] In one embodiment, when the processor implements the method of determining the target assessment result based on the result fusion strategy according to the first assessment result and / or the second assessment result, it is configured to implement:

[0118] In a case where the level difference between the first assessment result and the second assessment result is less than a preset threshold, based on the result fusion strategy, determining the first assessment result or the second assessment result as the target assessment result;

[0119] In a case where the level difference between the first assessment result and the second assessment result is greater than or equal to a preset threshold, a median level of the first assessment result and the second assessment result is determined as the target assessment result.

[0120] In one embodiment, when the processor implements the step of inputting the target object into the classification head and the preset regression head respectively, and obtaining the first evaluation result output by the classification head and the second evaluation result output by the regression head, it is configured to implement:

[0121] Obtaining the confidence of each sub-image corresponding to the damaged vehicle image;

[0122] When the confidence is greater than a preset confidence threshold, the sub-image is determined as the target object and input into the classification head and the preset regression head.

[0123] In one embodiment, when the processor implements the step of inputting the target object into the classification head and the preset regression head respectively, and obtaining the first evaluation result output by the classification head and the second evaluation result output by the regression head, it is configured to implement:

[0124] The regression prediction result of the regression head is obtained, and the second evaluation result Y=sigmoid(x)*n*α+β is determined according to the following formula, wherein Y is the second evaluation result, x is the regression prediction result, n is the number of vehicle damage levels, α is the first hyperparameter, and β is the second hyperparameter.

[0125] In one embodiment, when implementing the vehicle damage level assessment method, the processor is used to implement:

[0126] Obtaining a first evaluation result and a second evaluation result of a preset sample image, and determining first loss information of the classification head and second loss information of the regression head according to the annotation information corresponding to the sample image and the first evaluation result and the second evaluation result;

[0127] determining target loss information according to the first loss information and the second loss information;

[0128] When the target loss information does not meet the preset conditions, the classification head is adjusted according to the first loss information, and the first hyperparameter and the second hyperparameter of the regression head are adjusted according to the second loss information until the target loss information meets the preset conditions.

[0129] In one embodiment, when the processor determines the target loss information according to the first loss information and the second loss information, it is configured to implement:

[0130] Determine first loss data according to the product of the first loss information and a first preset coefficient, and determine second loss data according to the product of the second loss information and a second preset coefficient;

[0131] The target loss information is determined according to the sum of the first loss data and the second loss data.

[0132] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described vehicle damage level assessment can refer to the corresponding process in the aforementioned vehicle damage level assessment control method embodiment, and will not be repeated here.

[0133] An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. The computer program includes program instructions. The method implemented when the program instructions are executed can refer to the various embodiments of the vehicle damage level assessment method of the present application.

[0134] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc., equipped on the computer device.

[0135] It should be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0136] It should also be understood that the term "and / or" used in the specification of this application and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations. It should be noted that, in this article, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "including a..." does not exclude the presence of other identical elements in the process, method, article or system including the element.

[0137] The serial numbers of the embodiments of the present application are for description only and do not represent the advantages and disadvantages of the embodiments. The above description is only a specific implementation mode of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for assessing the level of damage to a vehicle, characterized in that: The method comprises: Acquire an image of a damaged vehicle to be assessed, and input the damaged vehicle image into a preset target detection model, wherein the target detection model includes a recognition network for identifying a target object and a classification head for classifying the target object; Acquire at least one target object obtained by the recognition network identifying the damaged vehicle image; Input the target object into the classification head and the preset regression head respectively, and obtain a first evaluation result output by the classification head and a second evaluation result output by the regression head; The first assessment result and the second assessment result are fused to obtain a target assessment result of the vehicle damage level.

2. The method for assessing the vehicle damage level according to claim 1, characterized in that: The fusing of the first assessment result and the second assessment result to obtain the target assessment result of the vehicle damage level includes: Get the preset result fusion strategy; Based on the result fusion strategy, the target assessment result is determined according to the first assessment result and / or the second assessment result.

3. The method for assessing the vehicle damage level according to claim 2, characterized in that: The determining the target assessment result based on the result fusion strategy according to the first assessment result and / or the second assessment result includes: In a case where the level difference between the first assessment result and the second assessment result is less than a preset threshold, based on the result fusion strategy, determining the first assessment result or the second assessment result as the target assessment result; In a case where the level difference between the first assessment result and the second assessment result is greater than or equal to a preset threshold, a median level of the first assessment result and the second assessment result is determined as the target assessment result.

4. The method for assessing the vehicle damage level according to claim 1, characterized in that: The step of inputting the target object into the classification head and the preset regression head respectively, and obtaining a first evaluation result output by the classification head and a second evaluation result output by the regression head, comprises: Obtaining the confidence of each sub-image corresponding to the damaged vehicle image; When the confidence is greater than a preset confidence threshold, the sub-image is determined as the target object and input into the classification head and the preset regression head.

5. The method for assessing the vehicle damage level according to claim 1, characterized in that: The step of inputting the target object into the classification head and the preset regression head respectively, and obtaining a first evaluation result output by the classification head and a second evaluation result output by the regression head, comprises: The regression prediction result of the regression head is obtained, and the second evaluation result Y=sigmoid(x)*n*α+β is determined according to the following formula, wherein Y is the second evaluation result, x is the regression prediction result, n is the number of vehicle damage levels, α is the first hyperparameter, and β is the second hyperparameter.

6. The method for assessing the vehicle damage level according to claim 5, characterized in that: The method further comprises: Obtaining a first evaluation result and a second evaluation result of a preset sample image, and determining first loss information of the classification head and second loss information of the regression head according to the annotation information corresponding to the sample image and the first evaluation result and the second evaluation result; determining target loss information according to the first loss information and the second loss information; When the target loss information does not meet the preset conditions, the classification head is adjusted according to the first loss information, and the first hyperparameter and the second hyperparameter of the regression head are adjusted according to the second loss information until the target loss information meets the preset conditions.

7. The method for assessing the vehicle damage level according to claim 6, characterized in that: The determining target loss information according to the first loss information and the second loss information includes: Determine first loss data according to the product of the first loss information and a first preset coefficient, and determine second loss data according to the product of the second loss information and a second preset coefficient; The target loss information is determined according to the sum of the first loss data and the second loss data.

8. A vehicle damage level assessment device, characterized in that: The vehicle damage level assessment device comprises: An image acquisition module, used for acquiring an image of a damaged vehicle to be assessed, and inputting the damaged vehicle image into a preset target detection model, wherein the target detection model includes a recognition network for identifying a target object and a classification head for classifying the target object; A target object determination module, used to obtain at least one target object obtained by the recognition network identifying the damaged vehicle image; An evaluation result determination module, used to input the target object into the classification head and the preset regression head respectively, and obtain a first evaluation result output by the classification head and a second evaluation result output by the regression head; The result fusion module is used to fuse the first assessment result and the second assessment result to obtain a target assessment result of the vehicle damage level.

9. A computer device, characterized in that: The computer device includes a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, the steps of the vehicle damage level assessment method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the vehicle damage level assessment method according to any one of claims 1 to 7 are implemented.

Images