CN114972229A - Loss assessment detection method and device based on material type, electronic equipment and medium - Google Patents
Loss assessment detection method and device based on material type, electronic equipment and medium Download PDFInfo
- Publication number
- CN114972229A CN114972229A CN202210536714.9A CN202210536714A CN114972229A CN 114972229 A CN114972229 A CN 114972229A CN 202210536714 A CN202210536714 A CN 202210536714A CN 114972229 A CN114972229 A CN 114972229A
- Authority
- CN
- China
- Prior art keywords
- damage
- type
- detection
- loss
- determining
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000000463 material Substances 0.000 title claims abstract description 228
- 238000001514 detection method Methods 0.000 title claims abstract description 127
- 238000000034 method Methods 0.000 claims abstract description 55
- 238000004364 calculation method Methods 0.000 claims abstract description 31
- 238000012549 training Methods 0.000 claims description 89
- 238000012545 processing Methods 0.000 claims description 35
- 230000006735 deficit Effects 0.000 claims description 13
- 238000012360 testing method Methods 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 11
- 231100000809 Damage Assessment Model Toxicity 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 5
- 238000013473 artificial intelligence Methods 0.000 abstract description 10
- 238000005516 engineering process Methods 0.000 description 18
- 238000010586 diagram Methods 0.000 description 11
- 230000008569 process Effects 0.000 description 10
- 238000004422 calculation algorithm Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 8
- 238000004891 communication Methods 0.000 description 7
- 238000010801 machine learning Methods 0.000 description 5
- 241000282414 Homo sapiens Species 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 239000002184 metal Substances 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000001939 inductive effect Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
- 230000007723 transport mechanism Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Medical Informatics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the field of artificial intelligence, and provides a loss assessment detection method and device based on material types, electronic equipment and a medium, wherein the method comprises the following steps: carrying out damage assessment identification on the damage image through a damage assessment detection model to obtain detection damage information including detection damage types and detection damage materials; when the detected damage material belongs to the confusion type, determining that the target loss weight is a first loss weight, otherwise, determining that the target loss weight is the first loss weight, wherein the detected damage material belonging to the confusion type has at least one preset associated material, and the first loss weight is larger than the second loss weight; and determining a target damage assessment result according to the target loss weight and the loss calculation result of the detected damaged material. According to the technical scheme of the embodiment, larger loss weight can be adopted for the detection damage material of the confusion type, so that the damage assessment detection model can identify damage according to more characteristics which are not easy to confuse, and the identification accuracy of the loss model is effectively improved.
Description
Technical Field
The invention belongs to the technical field of artificial intelligence, and particularly relates to a loss assessment detection method and device based on material types, electronic equipment and a medium.
Background
At present, the automobile loss assessment is gradually converted into intelligent loss assessment from traditional manual work, after a picture of a damaged part is shot, the damaged degree and the maintenance mode of the automobile are identified through a trained loss assessment detection model, a loss assessment result is automatically obtained, and the work efficiency of the loss assessment is greatly improved. However, with the development of the automobile manufacturing level, the materials on the automobile body are more and more, so that the damage types to be covered by the damage assessment detection model are more and more, the damage assessment detection model usually adopts a target detection technology, the detection precision is reduced along with the increase of the damage types, the materials are easily identified by mistake into similar materials, and the accuracy of subsequent damage assessment is affected.
Disclosure of Invention
The following is a summary of the subject matter described in detail herein. This summary is not intended to limit the scope of the claims.
The embodiment of the invention provides a loss assessment detection method, a loss assessment detection device, electronic equipment and a medium based on material types, which can improve the material detection precision of a loss assessment detection model and ensure the accuracy of a loss assessment result.
In a first aspect, an embodiment of the present invention provides a damage assessment method based on a material type, including:
acquiring a damage image, and inputting the damage image into a pre-trained damage assessment detection model;
carrying out damage assessment identification on the damage image through the damage assessment detection model to obtain detection damage information, wherein the detection damage information comprises a detection damage type and a detection damage material;
determining material types of the detection damage materials, wherein the material types comprise a confusion type and a conventional type, at least one preset associated material exists in the detection damage materials belonging to the confusion type, and the preset associated material does not exist in the detection damage materials belonging to the conventional type;
determining a target loss weight, wherein the target loss weight is a first loss weight when the detected damage material belongs to the confusion type, or the target loss weight is a second loss weight when the detected damage material belongs to the regular type, and the first loss weight is larger than the second loss weight;
performing loss calculation according to the target loss weight and the detected damage material, and determining the target damage material according to the result of the loss calculation;
and determining a damage assessment result according to the detection damage type and the target damage material.
In some embodiments, prior to said acquiring the lesion image, the method further comprises:
obtaining a plurality of training samples, wherein each training sample is labeled with labeled damage information in advance, and the labeled damage information comprises a labeled damage type and a labeled damage material;
and training the damage assessment detection model according to a plurality of training samples.
In some embodiments, said training said impairment detection model based on a plurality of said training samples comprises:
obtaining predicted damage information corresponding to each training sample through the damage assessment detection model, wherein the predicted damage information comprises a predicted damage type and a predicted damage material;
determining a confusion sample from the training samples, and determining the remaining training samples as conventional samples, wherein the marking damage type of the confusion sample is the same as the predicted damage type, and the predicted damage material of the confusion sample is a preset associated material of the marking damage material;
and performing loss calculation on the confusion sample according to the first loss weight, and performing loss calculation on the conventional sample according to the second loss weight so as to finish the training of the damage assessment detection model.
In some embodiments, the determining a confounding sample from the training samples comprises:
obtaining an alternative sample from the training sample;
when the marked damage type of the alternative sample is the same as the predicted damage type, acquiring a preset associated material table, wherein the associated material table is marked with the preset associated material corresponding to each marked damage material;
and determining a target associated material corresponding to the candidate sample according to the associated material table, and determining the candidate sample as the aliasing sample when the target associated material is the same as the predicted damage material of the candidate sample.
In some embodiments, the impairment prediction model comprises a ResNet50 network, and before obtaining the predicted impairment information corresponding to each of the training samples by the impairment detection model, the method further comprises:
selecting a plurality of test samples from the plurality of training samples and inputting the plurality of test samples to the ResNet50 network;
obtaining a feature map output by the ResNet50 network for each test sample;
determining brightness information of each feature of the feature map, and determining the feature of which the brightness information is greater than a preset threshold value as a key feature;
and determining a network layer used for extracting the key features in the ResNet50 network as a key processing layer.
In some embodiments, the obtaining, by the impairment detection model, predicted impairment information corresponding to each of the training samples includes:
inputting a plurality of said training samples to said ResNet50 network;
acquiring a plurality of key feature sequences obtained by the key processing layer by performing feature extraction processing on each training sample;
performing feature replacement processing on the plurality of key feature sequences to obtain a plurality of target feature sequences;
and inputting the target feature sequences corresponding to the training samples into a classifier of the ResNet50 network to obtain the predicted damage information corresponding to the training samples.
In some embodiments, the performing the feature replacement process on the plurality of key feature sequences includes:
determining at least one feature to be replaced in each key feature sequence;
and replacing the feature to be replaced of one key feature sequence into another key feature sequence until the feature to be replaced of all the key feature sequences is replaced.
In a second aspect, an embodiment of the present invention provides a damage assessment apparatus based on material types, including:
the image acquisition unit is used for acquiring a damage image and inputting the damage image into a pre-trained damage assessment detection model;
the identification unit is used for identifying the damage of the damage image through the damage assessment detection model to obtain detection damage information, and the detection damage information comprises a detection damage type and a detection damage material;
a material determining unit, configured to determine a material type of the detected damaged material, where the material type includes an obfuscated type and a conventional type, at least one preset associated material exists in the detected damaged material belonging to the obfuscated type, and the preset associated material does not exist in the detected damaged material belonging to the conventional type;
a loss weight determining unit, configured to determine a target loss weight, where the target loss weight is a first loss weight when the detected damage material belongs to the confusion type, or a second loss weight when the detected damage material belongs to the regular type, and the first loss weight is greater than the second loss weight;
the loss calculation unit is used for performing loss calculation according to the target loss weight and the detected damage material and determining the target damage material according to the result of the loss calculation;
and the damage assessment unit is used for determining a damage assessment result according to the detection damage type and the target damage material.
In a third aspect, an embodiment of the present invention provides an electronic device, including: a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method for detecting impairment based on material type according to the first aspect when executing the computer program.
In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, which stores a computer program for executing the method for detecting damage based on material type according to the first aspect.
The embodiment of the invention comprises the following steps: acquiring a damage image, and inputting the damage image into a pre-trained damage assessment detection model; carrying out damage assessment identification on the damage image through the damage assessment detection model to obtain detection damage information, wherein the detection damage information comprises a detection damage type and a detection damage material; determining material types of the detection damage materials, wherein the material types comprise an confusion type and a conventional type, at least one preset associated material exists in the detection damage materials belonging to the confusion type, and the preset associated material does not exist in the detection damage materials belonging to the conventional type; determining a target loss weight, wherein the target loss weight is a first loss weight when the detected damage material belongs to the confusion type, or the target loss weight is a second loss weight when the detected damage material belongs to the regular type, and the first loss weight is larger than the second loss weight; performing loss calculation according to the target loss weight and the detected damage material, and determining the target damage material according to the result of the loss calculation; and determining a damage assessment result according to the detection damage type and the target damage material. According to the technical scheme of the embodiment, larger loss weight can be adopted for detection damage information of the confusion type, so that the contribution degree of the confusable material to the identification of the damage assessment detection model is reduced, the damage assessment detection model can carry out damage assessment identification according to more confusable features, and the identification accuracy of the loss model is effectively improved under the condition of more materials.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
Drawings
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the example serve to explain the principles of the invention and not to limit the invention.
FIG. 1 is a flowchart of a method for damage assessment based on material type according to an embodiment of the present invention;
FIG. 2 is a flowchart of obtaining training samples of a damage assessment model according to another embodiment of the present invention;
FIG. 3 is a flow chart of training the impairment detection model according to another embodiment of the present invention;
FIG. 4 is a flow chart of determining aliased samples provided by another embodiment of the invention;
FIG. 5 is a flow diagram for determining a critical processing layer provided by another embodiment of the present invention;
fig. 6 is a schematic diagram of a ResNet50 network according to another embodiment of the present invention;
FIG. 7 is a flow chart for determining predicted damage information provided by another embodiment of the present invention;
FIG. 8 is a flow chart of feature replacement provided by another embodiment of the present invention;
FIG. 9 is a schematic diagram of an alternative feature provided by another embodiment of the present invention;
fig. 10 is a structural diagram of a damage assessment apparatus based on material type according to another embodiment of the present invention;
fig. 11 is a device diagram of an electronic apparatus according to another embodiment of the present invention. .
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
It should be noted that although functional blocks are partitioned in a schematic diagram of an apparatus and a logical order is shown in a flowchart, in some cases, the steps shown or described may be performed in a different order than the partitioning of blocks in the apparatus or the order in the flowchart. The terms "first," "second," and the like in the description, in the claims, or in the drawings described above, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.
The invention provides a loss assessment detection method, a loss assessment detection device, electronic equipment and a loss assessment medium based on material types, wherein the method comprises the following steps: carrying out damage assessment identification on the damage image through the damage assessment detection model to obtain detection damage information, wherein the detection damage information comprises a detection damage type and a detection damage material; determining material types of the detection damage materials, wherein the material types comprise an confusion type and a conventional type, at least one preset associated material exists in the detection damage materials belonging to the confusion type, and the preset associated material does not exist in the detection damage materials belonging to the conventional type; determining a target loss weight, wherein the target loss weight is a first loss weight when the detected damage material belongs to the confusion type, or the target loss weight is a second loss weight when the detected damage material belongs to the regular type, and the first loss weight is greater than the second loss weight; performing loss calculation according to the target loss weight and the detected damage material, and determining the target damage material according to the result of the loss calculation; and determining a damage assessment result according to the detection damage type and the target damage material. According to the technical scheme of this embodiment, can adopt bigger loss weight to the detection damage material of the type that confuses to reduce the contribution degree of the material that easily confuses to the identification of the loss assessment detection model, make the loss assessment detection model can carry out the loss assessment identification according to more characteristics that do not easily confuse, effectively improve the identification accuracy of loss model under the more condition of material.
The embodiment of the application can compile, acquire and process related data based on an artificial intelligence technology. Among them, Artificial Intelligence (AI) is a theory, method, technique and application device that simulates, extends and expands human Intelligence using a digital computer or a machine controlled by a digital computer, senses the environment, acquires knowledge and uses the knowledge to obtain the best result.
The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction devices, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a robot technology, a biological recognition technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.
Machine Learning (ML) is a multi-domain cross discipline, and relates to a plurality of disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. The special research on how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve the performance of the computer. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied to all fields of artificial intelligence. Machine learning and deep learning generally include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and formal education learning.
Computer Vision technology (CV) is a science for researching how to make a machine "see", and further refers to that a camera and a Computer are used to replace human eyes to perform machine Vision such as identification, tracking and measurement on a target, and further image processing is performed, so that the Computer processing becomes an image more suitable for human eyes to observe or is transmitted to an instrument to detect. As a scientific discipline, computer vision research-related theories and techniques attempt to build artificial intelligence systems that can capture information from images or multidimensional data. The computer vision technology generally includes image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content/behavior recognition, three-dimensional object reconstruction, 3D technology, virtual reality, augmented reality, synchronous positioning and map construction, automatic driving, intelligent transportation and other technologies, and also includes common biometric identification technologies such as face recognition and fingerprint recognition.
The target detection algorithm is an algorithm for detecting a target from an image, and is implemented by a two-stage (two-stage) target detection algorithm such as a region algorithm (R-CNN) based on a convolutional neural network feature, a fast region algorithm (fast R-CNN) based on a convolutional neural network feature, or may be implemented by a one-stage (one-stage) target detection algorithm such as a You Only Local One (YOLO) target detection algorithm, a Single Shot multi-box Detector (SSD).
It should be noted that the model in the embodiment of the present invention may be stored in a server, and the server may be an independent server, or may be a cloud server that provides basic cloud computing services such as cloud service, a cloud database, cloud computing, a cloud function, cloud storage, web service, cloud communication, middleware service, domain name service, security service, Content Delivery Network (CDN), big data, and an artificial intelligence platform.
As shown in fig. 1, fig. 1 is a flowchart of a method for detecting damage based on material type according to an embodiment of the present invention, where the method for detecting damage based on material type includes, but is not limited to, the following steps:
step S110, obtaining a damage image, and inputting the damage image into a pre-trained damage assessment detection model;
step S120, carrying out damage assessment identification on the damage image through a damage assessment detection model to obtain detection damage information, wherein the detection damage information comprises a detection damage type and a detection damage material;
step S130, determining material types of the detected damaged materials, wherein the material types comprise an confusion type and a conventional type, the detected damaged materials belonging to the confusion type have at least one preset associated material, and the detected damaged materials belonging to the conventional type do not have the preset associated material;
step S140, determining a target loss weight, wherein the target loss weight is a first loss weight when the detected damage material belongs to the confusion type, or the target loss weight is a second loss weight when the detected damage material belongs to the conventional type, and the first loss weight is greater than the second loss weight;
step S150, loss calculation is carried out according to the target loss weight and the detected damage material, and the target damage material is determined according to the result of the loss calculation;
and step S160, determining a damage assessment result according to the detected damage type and the target damage material.
It should be noted that the damage image may be shot by the user through the intelligent terminal for the vehicle, and uploaded to the server, and the damage assessment result is output through the damage assessment detection model deployed in the server, and of course, the damage image may also be obtained in a system import manner and input to the damage assessment detection model, and the obtaining manner of the damage image is not limited in this embodiment.
It should be noted that, in order to implement the damage assessment identification, the damage assessment detection model may use a target detection algorithm to detect the damaged part, and determine the detected damage type and the detected damage material by image identification after determining the damaged area, wherein the identification accuracy of the detected damage type is usually high, and will not cause too much influence on the damage assessment result, and the detected damage material is likely to cause false identification, so that it is necessary to adjust the loss weight according to the material type, for example, when the damaged part is made of rubber or a frosted part, the similarity of the two parts at the image identification level is high, and the two parts are likely to be confused, so that the two materials can be preset as the preset associated material, and the first loss weight with a larger value is used to perform loss calculation in the identification process, so as to reduce the contribution of the material to material identification, and enable the damage assessment detection model to determine the target damage material by combining with more features, the accuracy of the damage assessment detection model is ensured in the use scenes with more and more materials. The preset associated material may be set in the configuration process of the damage assessment detection model, and the number of the preset associated material of each material may be any, which is not limited herein.
It should be noted that the normal type is a type without a preset associated material, that is, confusion with other materials is not caused, therefore, a default loss weight may be preset as the weight of the normal type, that is, the second loss weight, to ensure that the loss assessment detection model can normally complete identification, when it is determined that the detected material type is the confusion type, an additionally set first loss weight may be obtained, or the first loss weight may be calculated by the confusion coefficient and the second loss weight according to the preset confusion coefficient, and the specific manner may be determined according to actual requirements.
In addition, referring to fig. 2, in an embodiment, before performing step S110 of the embodiment shown in fig. 1, the following steps are further included, but not limited to:
step S210, obtaining a plurality of training samples, wherein each training sample is labeled with labeled damage information in advance, and the labeled damage information comprises a labeled damage type and a labeled damage material;
step S220, training the damage assessment detection model according to the plurality of training samples.
It should be noted that the marking damage information of the training sample may be marked through a marking frame, or may be marked through text information, and a specific marking mode may be selected according to an actual requirement, which is not limited in this embodiment. It should be noted that the quantity of the marked damage information of each training sample may be any, for example, if an image of one training sample contains more damages, the marked damage information may be respectively marked for each damage, and the damage type and the material of the damaged part may be marked, for example, the marked damage type may be a common automobile damage type such as a depression, a scratch, and the marked damage material may be a material commonly used in automobiles such as a sheet metal part, a rubber part, and a frosted part. It should be noted that the marking damage information may also be text information marked according to a fixed format, and the marking damage type and the marking damage material are obtained through simple text segmentation, for example, if the marking damage information is "sheet metal part scratch", the marking damage material may be segmented as a sheet metal part, and the marking damage type is scratch, and a specific marking mode may be selected according to an actual demand, which is not limited in this embodiment.
In addition, referring to fig. 3, in an embodiment, step S220 of the embodiment shown in fig. 2 further includes, but is not limited to, the following steps:
step S310, obtaining predicted damage information corresponding to each training sample through a damage assessment detection model, wherein the predicted damage information comprises a predicted damage type and a predicted damage material;
step S320, determining a confusion sample from the training samples, and determining the rest training samples as conventional samples, wherein the marking damage type of the confusion sample is the same as the predicted damage type, and the predicted damage material of the confusion sample is a preset associated material of the marking damage material;
and step S330, performing loss calculation on the confusion sample according to the first loss weight, and performing loss calculation on the conventional sample according to the second loss weight to finish the training of the loss assessment detection model.
It should be noted that, in the training process of the damage assessment model, feature extraction needs to be performed from the training sample, prediction is performed according to the extracted features to obtain predicted damage information, and after loss calculation is performed according to the predicted damage information, a classifier is used for classifying to obtain a prediction result, therefore, the predicted damage information is not the final training result of the damage assessment model, but is intermediate information extracted from the network structure thereof, and a specific network layer for extracting the predicted damage information can be determined according to the network structure of the damage assessment model, for example, in the case that the damage assessment model adopts a ResNet50 network, information output by an intermediate recognition layer of the network carries a significant feature sequence, which can be extracted from the intermediate recognition layer as the predicted damage information, so that features with dominant characteristics can perform loss calculation through a larger loss weight, the contribution of the features with dominance characteristics in the training process is weakened, so that the confusion degree is reduced, and the accuracy of the damage assessment detection model is improved.
It should be noted that the prediction of the damage type may be determined by an image recognition technique, for example, if it is determined by the image recognition technique that there is a change in a plane in the image, it may be determined that the damage type is a sag; the damage material may be determined by an image recognition technology, for example, if the surface of the damaged portion is determined to have particles by the image recognition technology, the damaged portion may be determined to be a frosted part, and a specific damage prediction manner is not an improvement of this embodiment, and is not described herein again.
It should be noted that, because the accuracy of identifying the damage type is high, when the marked damage type is different from the predicted damage type, the training sample can be determined as a sample with an identification error, and a negative sample is used for subsequent training, and there is no confusion, so that the present embodiment does not discuss the situation with the identification error of the damage type too much.
It should be noted that, in order to improve the identification accuracy of the damage assessment model for the damaged material, after obtaining the damage prediction information, it may be determined whether each training sample is a confused sample, on the premise that the damage type is labeled the same as the predicted damage type, if the damage material is labeled the different from the predicted damage material, the training of the damage assessment model may not be converged, the prediction is wrong, or the model training is converged, but the two materials are easily confused, in order to eliminate the first case, the present embodiment sets the preset associated material for each material, when the damage material is labeled and the predicted damage material are mutually preset associated materials, it may be determined that the training sample is the confused sample, if the training sample is used for subsequent training, it is likely to cause the mistaken identification of the material by the damage assessment model in the using process, therefore, in order to improve the identification accuracy of the damage assessment model, after the confusion sample is determined, a first loss weight with a larger value is adopted for the confusion sample to weaken the characteristics of the confusion sample and reduce the contribution of the confusion sample to model training, so that the damage assessment detection model can learn more characteristics from the conventional sample, and the identification accuracy of the damage assessment detection model is improved.
In addition, referring to fig. 4, in an embodiment, step S320 of the embodiment shown in fig. 3 further includes, but is not limited to, the following steps:
step S410, obtaining alternative samples from training samples;
step S420, when the marked damage type of the alternative sample is the same as the predicted damage type, acquiring a preset associated material table, wherein the preset associated material corresponding to each marked damage material is recorded in the associated material table;
step S430, determining a target associated material corresponding to the candidate sample according to the associated material table, and determining the candidate sample as a mixed sample when the target associated material is the same as the predicted damage material of the candidate sample.
It should be noted that, as the types of materials used by the automobile are more and more, the preset associated material is determined by importing the associated material table in the embodiment, so that the accuracy of the damage assessment detection model for identifying the automobile material can be effectively improved. Of course, after the associated material table is obtained and the training samples are obtained, the target associated material corresponding to each training sample may be added to the labeling damage information, so as to improve the efficiency of determining the confusing sample.
It should be noted that, because there are many training samples, the confusion sample may be determined by traversing the training samples, for example, each training sample is sequentially determined as an alternative sample, and the confusion sample is determined for the alternative sample, so that the comprehensiveness of the confusion sample determination can be ensured.
It should be noted that, after the candidate sample is determined, a table lookup operation may be performed from the associated material table according to the marked damage material of the candidate sample to determine a target associated material, where the target associated material may be multiple, and the specific number is not limited in this embodiment.
In addition, in an embodiment, the damage prediction model includes a ResNet50 network, and referring to fig. 5, before performing step S310 of the embodiment shown in fig. 3, the following steps are included, but not limited to:
step S510, selecting a plurality of test samples from a plurality of training samples, and inputting the plurality of test samples into a ResNet50 network;
step S520, obtaining a characteristic diagram output by the ResNet50 network aiming at each test sample;
step S530, determining the brightness information of each feature of the feature map, and determining the feature with the brightness information being larger than a preset threshold value as a key feature;
and step S540, determining a network layer used for extracting the key features in the ResNet50 network as a key processing layer.
It should be noted that the ResNet50 network is a commonly used classification network, and a network structure thereof may refer to fig. 6, of course, fig. 6 is only an example of a network structure, and is not a limitation on a specific network structure, and a person skilled in the art may adjust the network structure according to an actual situation, which is not described herein in detail.
It is worth noting that in order to determine the predicted damage information, a key processing layer in the ResNet50 network needs to be determined first, and the output of the key processing layer is used as the predicted damage information, so that the predicted damage information can carry more key features, the predicted damage material of the training sample is more accurate, and the confidence of the confusing sample is improved.
It should be noted that the key processing layer of the ResNet50 network is usually fixed, and therefore can be determined by a test sample during first training and directly used in a subsequent use process, which is not limited in this embodiment. It should be noted that, in order to determine the key processing layer, a plurality of test samples may be obtained from the training samples, and of course, in order to improve the accuracy of the feature map, all the training samples may also be used as the test samples, and may be selected according to actual requirements.
It is to be noted that after the feature extraction is performed on the test sample, an area with more features is usually a key area, and since there are more features, the area is highlighted in the visualized feature map, that is, the luminance information is higher, based on this, the key features may be screened by setting a preset threshold, so as to determine the neurons with dominating features, for example, for the network structure shown in fig. 6, it is determined that the extracted network layer is the identification block 3x according to the key features, and then the identification block 3x may be determined as a key processing layer, and the output of the key processing layer is obtained as the predicted damage information.
In addition, referring to fig. 7, in an embodiment, step S310 of the embodiment shown in fig. 3 further includes, but is not limited to, the following steps:
step S710, inputting a plurality of training samples into a ResNet50 network;
step S720, acquiring a plurality of key characteristic sequences obtained by the key processing layer by performing characteristic extraction processing on each training sample;
step S730, performing feature replacement processing on the plurality of key feature sequences to obtain a plurality of target feature sequences;
step S740, inputting the target feature sequences corresponding to the training samples into the classifier of the ResNet50 network, and obtaining the predicted damage information corresponding to the training samples.
It should be noted that, in order to enable the damage assessment detection model to perform damage identification according to more features, it is necessary to weaken the contribution of the key features to the damage identification to some extent, and in addition to applying a larger loss weight to the aliasing sample, the key features having a key role may be replaced, so as to suppress neurons having dominance traits, and make the determined aliasing sample have a higher confidence.
It should be noted that, after the key processing layer is determined, a key feature sequence output by the key processing layer for each training sample may be obtained, where each key feature sequence includes a key feature of a corresponding training sample, and since the key feature is for a specific training sample, that is, the key feature of the key feature sequence 1 does not form a key contribution in the key feature sequence 2, the contribution of the key feature to the prediction result may be weakened by performing feature replacement between multiple key feature sequences, in this case, if the training sample is still determined to be a confusing sample, it may be determined that the material of the confusing sample belongs to a confusing material, so that loss calculation is performed according to a larger loss weight, and the identification accuracy of the damage assessment detection model is improved.
It should be noted that the key feature sequence generally has a plurality of features, and feature replacement may be performed in any form, so as to ensure that the replaced features are derived from other key feature sequences, for example, as shown in fig. 9, in the case of having 3 key feature sequences, the features of the key feature sequence 1 may be replaced according to the features of the key feature sequence 2, or may be replaced by the features of the key feature sequence 3, or a plurality of features may be acquired from each of the key feature sequence 2 and the key feature sequence 3 at the same time for replacement, and a specific replacement manner may be selected according to actual needs, which is not limited in this embodiment.
It should be noted that the classifier of the ResNet50 network may be a full link layer in the network structure shown in fig. 6, and after obtaining the prediction result obtained by the loss calculation, the classifier performs classification according to features to determine the prediction result, which is a process well known to those skilled in the art and is not repeated herein for simplicity of description.
In addition, referring to fig. 8, in an embodiment, step S830 of the embodiment shown in fig. 7 further includes, but is not limited to, the following steps:
step S810, determining at least one feature to be replaced in each key feature sequence;
step S820, replace the feature to be replaced of one key feature sequence into another key feature sequence until the feature to be replaced of all the key feature sequences is replaced.
It should be noted that, the feature replacement in this embodiment may adopt a random replacement mode, that is, N features to be replaced are selected from one key feature sequence, and N features to be replaced are obtained from other key feature sequences for replacement, in order to better explain the technical solution of this embodiment, a specific example is proposed below with reference to fig. 9:
in this example, the loss assessment detection model includes 3 training samples, and takes a ResNet50 network shown in fig. 6 as an example, after 3 training samples are input to a ResNet50 network, a recognition block 3x is taken as a key processing layer, 3 key feature sequences output by the recognition block are obtained as shown in fig. 9, each key feature sequence includes 5 features, and 2 features to be replaced are determined from each key feature sequence, which are respectively features 1-2 and 1-4 in a key feature sequence 1, features 2-2 and 2-4 in a key feature sequence 2, and features 3-2 and 3-4 in a key feature sequence 3; replacing the features 2-2 and 2-4 in the key feature sequence 2 with the key feature sequence 3 to obtain a target feature sequence 3, replacing the features 3-2 and 3-4 in the key feature sequence 3 with the key feature sequence 1 to obtain a target feature sequence 1, and replacing the features 1-2 and 1-4 in the key feature sequence 1 with the key feature sequence 2 to obtain a target feature sequence 2.
It should be noted that the feature to be replaced of the key feature sequence may be replaced with any other key feature sequence, so as to ensure that the feature quantity of each key feature sequence after the feature replacement processing is consistent, and each key feature sequence is subjected to the feature replacement processing to obtain the target feature sequence.
It should be noted that, since the key feature sequences are in one-to-one correspondence with the training samples, the number of the key feature sequences is the same as the number of the training samples, in order to improve the training efficiency, the number of the replacement sequences may be preset, a plurality of target replacement feature sequences that need to be subjected to feature liking processing are screened from all the training samples, a specific numerical value of the number of the replacement sequences may be adjusted according to an actual requirement, of course, all the key feature sequences may also be selected as the target replacement feature sequences, which is not limited in this embodiment.
In addition, referring to fig. 10, an embodiment of the present invention provides a damage assessment apparatus based on a material type, where the damage assessment apparatus 1000 based on a material type includes:
an image acquisition unit 1010, configured to acquire a damage image and input the damage image to a pre-trained damage assessment detection model;
the identification unit 1020 is configured to perform damage assessment identification on the damage image through a damage assessment detection model to obtain detected damage information, where the detected damage information includes a detected damage type and a detected damage material;
a material determining unit 1030, configured to determine a material type of the detected damaged material, where the material type includes an obfuscated type and a conventional type, and the detected damaged material belonging to the obfuscated type has at least one preset associated material, and the detected damaged material belonging to the conventional type does not have a preset associated material;
the loss weight determining unit 1040 is configured to determine a target loss weight, where the target loss weight is a first loss weight when the detected damage material belongs to the confusion type, or the target loss weight is a second loss weight when the detected damage material belongs to the regular type, and the first loss weight is greater than the second loss weight;
a loss calculation unit 1050, configured to perform loss calculation according to the target loss weight and the detected damaged material, and determine the target damaged material according to a result of the loss calculation;
and a damage assessment unit 1060, configured to determine a damage assessment result according to the detected damage type and the target damage material.
In addition, referring to fig. 11, an embodiment of the present invention also provides an electronic device 1100, including: memory 1110, processor 1120, and computer programs stored on memory 1110 and executable on processor 1120.
The processor 1120 and the memory 1110 may be connected by a bus or other means.
The non-transitory software programs and instructions required to implement the damage detection method based on the material type of the above embodiment are stored in the memory 1110, and when executed by the processor 1120, the damage detection method based on the material type of the above embodiment is performed, for example, the method steps S110 to S160 in fig. 1, the method steps S210 to S220 in fig. 2, the method steps S310 to S330 in fig. 3, the method steps S410 to S430 in fig. 4, the method steps S510 to S540 in fig. 5, the method steps S710 to S740 in fig. 7, and the method steps S810 to S820 in fig. 8, which are described above, are performed.
The above described embodiments of the apparatus are merely illustrative, wherein the units illustrated as separate components may or may not be physically separate, may be located in one place, or may be distributed over a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment.
Furthermore, an embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored, and the computer program is executed by a processor or a controller, for example, by a processor in the above embodiment of the electronic device, so that the processor executes the method for detecting damage based on material type in the above embodiment, for example, execute the above-described method steps S110 to S160 in fig. 1, method steps S210 to S220 in fig. 2, method steps S310 to S330 in fig. 3, method steps S410 to S430 in fig. 4, method steps S510 to S540 in fig. 5, method steps S710 to S740 in fig. 7, and method steps S810 to S820 in fig. 8. It will be understood by those of ordinary skill in the art that all or some of the steps, means, and methods disclosed above may be implemented as software, firmware, hardware, or suitable combinations thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable storage media, which may include computer storage media (or non-transitory storage media) and communication storage media (or transitory storage media). The term computer storage media includes volatile and nonvolatile, removable and non-removable storage media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as is well known to those of ordinary skill in the art. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, Digital Versatile Disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other storage medium which can be used to store the desired information and which can be accessed by a computer. In addition, communication storage media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery storage media as is well known to those of ordinary skill in the art.
The embodiments are operational with numerous general purpose or special purpose computing device environments or configurations. For example: personal computers, server computers, hand-held or portable electronic devices, tablet electronic devices, multiprocessor apparatus, microprocessor-based apparatus, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above devices or electronic devices, and the like. The application may be described in the general context of computer programs, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing electronic devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage electronic devices.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present application. Each block in the flowchart or block diagrams may represent a module, a segment, or a portion of code, which comprises one or more programs for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based apparatus that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The units described in the embodiments of the present application may be implemented by software, or may be implemented by hardware, and the described units may also be disposed in a processor. Wherein the names of the elements do not in some way constitute a limitation on the elements themselves.
It should be noted that although in the above detailed description several modules or units of the electronic device for action execution are mentioned, this division is not mandatory. Indeed, the features and functionality of two or more modules or units described above may be embodied in one module or unit, according to embodiments of the application. Conversely, the features and functions of one module or unit described above may be further divided into embodiments by a plurality of modules or units.
Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein may be implemented by software, and may also be implemented by software in combination with necessary hardware. Therefore, the technical solution according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a usb disk, a removable hard disk, etc.) or on a network, and includes several instructions to enable a computing electronic device (which can be a personal computer, a server, a touch terminal, or a network electronic device, etc.) to execute the method according to the embodiments of the present application.
The electronic device of the present embodiment may include: radio Frequency (RF) circuit, memory, input unit, display unit, sensor, audio circuit, wireless fidelity (WiFi) module, processor, and power supply. The RF circuit can be used for receiving and transmitting signals in the process of information receiving and transmitting or conversation, and particularly, the downlink information of the base station is received and then is processed by the processor; in addition, data for designing uplink is transmitted to the base station. Typically, the RF circuitry includes, but is not limited to, an antenna, at least one Amplifier, a transceiver, a coupler, a Low Noise Amplifier (LNA), a duplexer, and the like. In addition, the RF circuitry may also communicate with networks and other devices via wireless communications. The wireless communication may use any communication standard or protocol, including but not limited to Global System for Mobile communication (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, Short Messaging Service (SMS), and the like. The memory may be used to store software programs and modules, and the processor may execute various functional applications and data processing of the electronic device by operating the software programs and modules stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data (such as audio data, a phonebook, etc.) created according to the use of the electronic device, and the like.
Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains.
It will be understood that the present application is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the application is limited only by the appended claims.
While the preferred embodiments of the present invention have been described, the present invention is not limited to the above embodiments, and those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and such equivalent modifications or substitutions are to be included within the scope of the present invention defined by the appended claims.
Claims (10)
1. A loss assessment detection method based on material types is characterized by comprising the following steps:
acquiring a damage image, and inputting the damage image into a pre-trained damage assessment detection model;
carrying out damage assessment identification on the damage image through the damage assessment detection model to obtain detection damage information, wherein the detection damage information comprises a detection damage type and a detection damage material;
determining material types of the detection damage materials, wherein the material types comprise an confusion type and a conventional type, at least one preset associated material exists in the detection damage materials belonging to the confusion type, and the preset associated material does not exist in the detection damage materials belonging to the conventional type;
determining a target loss weight, wherein the target loss weight is a first loss weight when the detected damage material belongs to the confusion type, or the target loss weight is a second loss weight when the detected damage material belongs to the regular type, and the first loss weight is greater than the second loss weight;
performing loss calculation according to the target loss weight and the detected damage material, and determining the target damage material according to the result of the loss calculation;
and determining a damage assessment result according to the detection damage type and the target damage material.
2. The method of claim 1, wherein before the obtaining the damage image, the method further comprises:
obtaining a plurality of training samples, wherein each training sample is labeled with labeled damage information in advance, and the labeled damage information comprises a labeled damage type and a labeled damage material;
and training the damage assessment detection model according to a plurality of training samples.
3. The method of claim 2, wherein the training the impairment detection model based on the plurality of training samples comprises:
obtaining predicted damage information corresponding to each training sample through the damage assessment detection model, wherein the predicted damage information comprises a predicted damage type and a predicted damage material;
determining a confusion sample from the training samples, and determining the remaining training samples as conventional samples, wherein the marking damage type of the confusion sample is the same as the predicted damage type, and the predicted damage material of the confusion sample is the preset associated material of the marking damage material;
and performing loss calculation on the confusion sample according to the first loss weight, and performing loss calculation on the conventional sample according to the second loss weight so as to finish the training of the damage assessment detection model.
4. The method according to claim 3, wherein the determining the aliasing sample from the training sample comprises:
obtaining an alternative sample from the training sample;
when the marked damage type of the alternative sample is the same as the predicted damage type, acquiring a preset associated material table, wherein the associated material table is marked with the preset associated material corresponding to each marked damage material;
and determining a target associated material corresponding to the candidate sample according to the associated material table, and determining the candidate sample as the aliasing sample when the target associated material is the same as the predicted damage material of the candidate sample.
5. The method of claim 3, wherein the impairment prediction model comprises a ResNet50 network, and before obtaining the predicted impairment information corresponding to each of the training samples by the impairment detection model, the method further comprises:
selecting a plurality of test samples from the plurality of training samples and inputting the plurality of test samples to the ResNet50 network;
obtaining a feature map output by the ResNet50 network for each test sample;
determining brightness information of each feature of the feature map, and determining the feature of which the brightness information is greater than a preset threshold value as a key feature;
and determining a network layer used for extracting the key features in the ResNet50 network as a key processing layer.
6. The method according to claim 3, wherein the obtaining of the predicted damage information corresponding to each of the training samples by the damage assessment model comprises:
inputting a plurality of said training samples to said ResNet50 network;
acquiring a plurality of key feature sequences obtained by the key processing layer by performing feature extraction processing on each training sample;
performing feature replacement processing on the plurality of key feature sequences to obtain a plurality of target feature sequences;
and inputting the target feature sequences corresponding to the training samples into a classifier of the ResNet50 network to obtain the predicted damage information corresponding to the training samples.
7. The method according to claim 6, wherein the performing the feature replacement processing on the plurality of key feature sequences comprises:
determining at least one feature to be replaced in each key feature sequence;
and replacing the feature to be replaced of one key feature sequence into another key feature sequence until the feature to be replaced of all the key feature sequences is replaced.
8. The utility model provides a loss assessment detection device based on material type which characterized in that includes:
the image acquisition unit is used for acquiring a damage image and inputting the damage image to a pre-trained damage assessment detection model;
the identification unit is used for identifying the damage of the damage image through the damage assessment detection model to obtain detection damage information, and the detection damage information comprises a detection damage type and a detection damage material;
a material determining unit, configured to determine a material type of the detected damaged material, where the material type includes an obfuscated type and a conventional type, at least one preset associated material exists in the detected damaged material belonging to the obfuscated type, and the preset associated material does not exist in the detected damaged material belonging to the conventional type;
a loss weight determining unit, configured to determine a target loss weight, where the target loss weight is a first loss weight when the detected damage material belongs to the confusion type, or a second loss weight when the detected damage material belongs to the regular type, and the first loss weight is greater than the second loss weight;
the loss calculation unit is used for performing loss calculation according to the target loss weight and the detected damage material and determining the target damage material according to the result of the loss calculation;
and the damage assessment unit is used for determining a damage assessment result according to the detection damage type and the target damage material.
9. An electronic device, comprising: memory, processor and computer program stored on the memory and executable on the processor, wherein the processor implements the method of impairment detection based on material type according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium storing a computer program for executing the method for detecting damage based on material type according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210536714.9A CN114972229A (en) | 2022-05-17 | 2022-05-17 | Loss assessment detection method and device based on material type, electronic equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210536714.9A CN114972229A (en) | 2022-05-17 | 2022-05-17 | Loss assessment detection method and device based on material type, electronic equipment and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114972229A true CN114972229A (en) | 2022-08-30 |
Family
ID=82983533
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210536714.9A Pending CN114972229A (en) | 2022-05-17 | 2022-05-17 | Loss assessment detection method and device based on material type, electronic equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114972229A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110570316A (en) * | 2018-08-31 | 2019-12-13 | 阿里巴巴集团控股有限公司 | method and device for training damage recognition model |
| US20200090321A1 (en) * | 2018-09-07 | 2020-03-19 | Alibaba Group Holding Limited | System and method for facilitating efficient damage assessments |
| CN111667011A (en) * | 2020-06-08 | 2020-09-15 | 平安科技（深圳）有限公司 | Damage detection model training method, damage detection model training device, damage detection method, damage detection device, damage detection equipment and damage detection medium |
| CN111839360A (en) * | 2020-06-22 | 2020-10-30 | 珠海格力电器股份有限公司 | Data processing method, device and equipment of sweeper and computer readable medium |
-
2022
- 2022-05-17 CN CN202210536714.9A patent/CN114972229A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110570316A (en) * | 2018-08-31 | 2019-12-13 | 阿里巴巴集团控股有限公司 | method and device for training damage recognition model |
| US20200090321A1 (en) * | 2018-09-07 | 2020-03-19 | Alibaba Group Holding Limited | System and method for facilitating efficient damage assessments |
| CN111667011A (en) * | 2020-06-08 | 2020-09-15 | 平安科技（深圳）有限公司 | Damage detection model training method, damage detection model training device, damage detection method, damage detection device, damage detection equipment and damage detection medium |
| CN111839360A (en) * | 2020-06-22 | 2020-10-30 | 珠海格力电器股份有限公司 | Data processing method, device and equipment of sweeper and computer readable medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110209866A (en) | A kind of image search method, device, equipment and computer readable storage medium | |
| CN111652087B (en) | Car inspection method, device, electronic equipment and storage medium | |
| KR20190106853A (en) | Apparatus and method for recognition of text information | |
| CN107808126A (en) | Vehicle retrieval method and device | |
| CN111931859B (en) | Multi-label image recognition method and device | |
| CN112215207A (en) | Remote sensing image airplane target detection method combining multi-scale and attention mechanism | |
| CN104615986A (en) | Method for utilizing multiple detectors to conduct pedestrian detection on video images of scene change | |
| CN111121797B (en) | Road screening method, device, server and storage medium | |
| CN110473211B (en) | Method and equipment for detecting number of spring pieces | |
| CN109857878B (en) | Article labeling method and device, electronic equipment and storage medium | |
| CN110609879B (en) | Interest point duplicate determination method and device, computer equipment and storage medium | |
| CN105426862A (en) | Analysis method and system based on RFID (Radio Frequency Identification Devices), positioning technology and video technology | |
| KR20200017611A (en) | Method and apparatus for positioning by using Deep learning | |
| CN115205883A (en) | Data auditing method, device, equipment and storage medium based on OCR (optical character recognition) and NLP (non-line language) | |
| KR20200017612A (en) | Method for positioning learning by using Deep learning | |
| CN115393606A (en) | Method and system for image recognition | |
| CN111126112B (en) | Candidate region determination method and device | |
| CN116152576B (en) | Image processing method, device, equipment and storage medium | |
| CN109726726B (en) | Event detection method and device in video | |
| CN114972229A (en) | Loss assessment detection method and device based on material type, electronic equipment and medium | |
| CN116958512A (en) | Target detection method, target detection device, computer readable medium and electronic equipment | |
| CN105512664A (en) | Image recognition method and device | |
| CN115984723A (en) | Road damage detection method, system, device, storage medium and computer equipment | |
| CN115393755A (en) | Visual target tracking method, device, equipment and storage medium | |
| CN112017634A (en) | Data processing method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |