CN112634245A - Loss detection model training method, loss detection method and device - Google Patents
Loss detection model training method, loss detection method and device Download PDFInfo
- Publication number
- CN112634245A CN112634245A CN202011586533.4A CN202011586533A CN112634245A CN 112634245 A CN112634245 A CN 112634245A CN 202011586533 A CN202011586533 A CN 202011586533A CN 112634245 A CN112634245 A CN 112634245A
- Authority
- CN
- China
- Prior art keywords
- loss
- appearance image
- detection model
- appearance
- loss detection
- 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
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
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a loss detection model training method, a loss detection method and a loss detection device. Further, a deep neural network is built using the appearance image dataset to train a loss detection model that can be used to detect appearance loss of the smart device. Based on the method, a loss detection model is trained in advance through a large number of intelligent device appearance images serving as experience data, when the loss degree requirement for detecting the intelligent device to be detected is generated, the loss degree of the intelligent device to be detected can be rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent device recovery.
Description
Technical Field
The invention relates to the technical field of electronic products, in particular to a loss detection model training method, a loss detection method and a loss detection device.
Background
With the development of electronic product technology, various intelligent devices such as smart phones, notebook computers, tablet computers, and the like are developed. At present, along with the rapid development of economy and technology, the popularization and the updating speed of intelligent equipment are also faster and faster. Taking a smart phone as an example, the coming of the 5G era accelerates the generation change of the smart phone. In the iterative process of the intelligent equipment, effective recovery is one of effective utilization means of the residual value of the intelligent equipment, and the chemical pollution to the environment and the waste can be reduced.
In the recovery process of the intelligent equipment, the overall loss degree of the intelligent equipment has great influence on the recovery evaluation of the intelligent equipment. Generally, the overall loss of the intelligent device is determined mainly by observing the appearance loss of the intelligent device, such as the appearance loss of the categories of scratches, dropped paint or outbreaks, so as to evaluate the overall loss of the intelligent device, and a part of effective reference is provided for the recovery evaluation of the intelligent device.
However, the conventional method for detecting the loss of the smart device mainly determines the loss and the degree of the loss according to the subjective judgment of a professional quality inspector through the observation of the human eyes of the professional quality inspector. However, human eye observation is time-consuming and labor-consuming, and subjective judgment is affected by factors such as experience and state of professional quality inspectors, so that the accuracy of judgment is difficult to ensure.
Disclosure of Invention
Therefore, it is necessary to provide a loss detection model training method, a loss detection method and a loss detection device for the defects of the conventional method for detecting the loss of the smart device.
A loss detection model training method comprises the following steps:
acquiring an appearance image of the intelligent device;
marking a loss category for the appearance image of the intelligent equipment to establish an appearance image data set;
and establishing a deep neural network by using the appearance image data set so as to train a loss detection model which can be used for detecting the appearance loss of the intelligent equipment.
According to the loss detection model training method, after the appearance image of the intelligent device is obtained, loss categories are marked for the appearance image of the intelligent device, so that an appearance image data set is established. Further, a deep neural network is built using the appearance image dataset to train a loss detection model that can be used to detect appearance loss of the smart device. Based on the method, a loss detection model is trained in advance through a large number of intelligent device appearance images serving as experience data, when the loss degree requirement for detecting the intelligent device to be detected is generated, the loss degree of the intelligent device to be detected can be rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent device recovery.
In one embodiment, the process of labeling the smart device appearance image with a loss category to create an appearance image dataset includes the steps of:
selecting an intelligent equipment appearance image of a data set to be established according to the proportion of the average loss categories;
and determining the proportion of a training set and a test set in the appearance image of the intelligent equipment with the data set to be established so as to determine the appearance image data set.
In one embodiment, the wear categories include scratches, paint drops, and pop-ups.
In one embodiment, the ratio of training set to test set is 9: 1.
In one embodiment, the deep neural network comprises a YOLO neural network.
In one embodiment, the process of building a deep neural network using the appearance image dataset to train a loss detection model for detecting loss classes of the appearance of the smart device includes the steps of:
determining the number of filters of the deep neural network according to the number of the loss categories;
modifying the image size in the appearance image data set to be a multiple of the step length to the power of the down-sampling times;
modifying impulse size, pooling algorithm, resetting loss function and learning rate;
and adding a prediction layer for detecting the appearance loss to determine a loss detection model with optimal weight.
In one embodiment, the YOLO neural network comprises a YOLOV3 neural network or a YOLOV5 neural network.
A wear detection model training apparatus, comprising:
the image acquisition module is used for acquiring an appearance image of the intelligent equipment;
the data set establishing module is used for marking the loss category for the appearance image of the intelligent equipment so as to establish an appearance image data set;
and the model training module is used for establishing a deep neural network by using the appearance image data set so as to train a loss detection model which can be used for detecting the appearance loss of the intelligent equipment.
After the appearance image of the intelligent device is obtained, the loss detection model training device marks the loss category for the appearance image of the intelligent device so as to establish an appearance image data set. Further, a deep neural network is built using the appearance image dataset to train a loss detection model that can be used to detect appearance loss of the smart device. Based on the method, a loss detection model is trained in advance through a large number of intelligent device appearance images serving as experience data, when the loss degree requirement for detecting the intelligent device to be detected is generated, the loss degree of the intelligent device to be detected can be rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent device recovery.
A computer storage medium having computer instructions stored thereon, the computer instructions when executed by a processor implementing the wear detection model training method of any of the above embodiments.
After the appearance image of the intelligent device is obtained, the computer storage medium marks the loss category for the appearance image of the intelligent device so as to establish an appearance image data set. Further, a deep neural network is built using the appearance image dataset to train a loss detection model that can be used to detect appearance loss of the smart device. Based on the method, a loss detection model is trained in advance through a large number of intelligent device appearance images serving as experience data, when the loss degree requirement for detecting the intelligent device to be detected is generated, the loss degree of the intelligent device to be detected can be rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent device recovery.
A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the wear detection model training method of any of the above embodiments when executing the program.
After the appearance image of the intelligent device is obtained, the computer device marks the loss category for the appearance image of the intelligent device so as to establish an appearance image data set. Further, a deep neural network is built using the appearance image dataset to train a loss detection model that can be used to detect appearance loss of the smart device. Based on the method, a loss detection model is trained in advance through a large number of intelligent device appearance images serving as experience data, when the loss degree requirement for detecting the intelligent device to be detected is generated, the loss degree of the intelligent device to be detected can be rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent device recovery.
A loss detection method comprising the steps of:
acquiring an appearance image to be detected; the appearance image to be detected is an appearance image of the intelligent equipment to be detected;
performing loss class probability prediction on the appearance image to be detected according to a loss detection model;
and determining the appearance loss degree of the intelligent device to be tested according to the loss category probability.
According to the loss detection method, after the appearance image to be detected is obtained, loss class probability prediction is carried out on the appearance image to be detected according to the loss detection model, and the appearance loss degree of the intelligent device to be detected is determined according to the loss class probability. Based on this, the loss degree of the intelligent equipment to be tested is rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent equipment recovery.
A wear detection apparatus comprising:
the to-be-detected image acquisition module is used for acquiring an appearance image to be detected; the appearance image to be detected is an appearance image of the intelligent equipment to be detected;
the probability prediction module is used for performing loss category probability prediction on the appearance image to be detected according to the loss detection model;
and the loss determining module is used for determining the loss degree of the intelligent equipment to be tested according to the loss category probability.
After the appearance image to be detected is obtained, the loss detection device predicts the loss class probability of the appearance image to be detected according to the loss detection model, and determines the appearance loss degree of the intelligent device to be detected according to the loss class probability. Based on this, the loss degree of the intelligent equipment to be tested is rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent equipment recovery.
A computer storage medium having stored thereon computer instructions which, when executed by a processor, implement the wear detection method of any of the above embodiments.
After the appearance image to be detected is obtained, the computer storage medium carries out loss category probability prediction on the appearance image to be detected according to the loss detection model, and determines the appearance loss degree of the intelligent device to be detected according to the loss category probability. Based on this, the loss degree of the intelligent equipment to be tested is rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent equipment recovery.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the wear detection method of any of the above embodiments when executing the program.
After the computer equipment obtains the appearance image to be detected, the loss class probability of the appearance image to be detected is predicted according to the loss detection model, and the appearance loss degree of the intelligent equipment to be detected is determined according to the loss class probability. Based on this, the loss degree of the intelligent equipment to be tested is rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent equipment recovery.
Drawings
FIG. 1 is a flow diagram of a loss detection model training method according to an embodiment;
FIG. 2 is a flow chart of a loss detection model training method according to another embodiment;
FIG. 3 is a flow chart of a loss detection model training method according to yet another embodiment;
FIG. 4 is a block diagram of a wear detection model training apparatus according to an embodiment;
FIG. 5 is a flow chart of a wear detection method according to one embodiment;
FIG. 6 is a flow chart of convolution of an appearance image to be measured in a loss detection model;
fig. 7 is a block diagram of a wear detection device according to an embodiment.
Detailed Description
For better understanding of the objects, technical solutions and effects of the present invention, the present invention will be further explained with reference to the accompanying drawings and examples. Meanwhile, the following described examples are only for explaining the present invention, and are not intended to limit the present invention.
The embodiment of the invention provides a loss detection model training method.
Fig. 1 is a flowchart of a loss detection model training method according to an embodiment, and as shown in fig. 1, the loss detection model training method according to an embodiment includes steps S100 to S102:
s100, acquiring an appearance image of the intelligent device;
the intelligent device comprises an intelligent mobile phone, a notebook computer, a tablet computer and the like. Taking a smart device as an example of a mobile phone, the degree of appearance loss of the mobile phone is generally evaluated by evaluating a large-area contact surface such as a front surface or a back surface of the mobile phone. Meanwhile, the front of the mobile phone is generally a screen occupying a large proportion area. Therefore, the back picture of the intelligent device can be acquired as the appearance image of the intelligent device.
In one embodiment, based on the priori knowledge, the loss part in the whole appearance image of the intelligent device is cut out, and the cut image is used as the appearance image of the intelligent device.
S101, marking loss categories for the appearance images of the intelligent equipment to establish appearance image data sets;
in one embodiment, the wear categories include a plurality of a priori knowledge based wear types including scratches, paint drops, implosions, and the like. It is essential that the categories of wear include, but are not limited to, scratches, paint drops, and pop-ups. On the premise that the loss detection model is determined, the loss classes can be adjusted to be increased or decreased through the loss classes, and the loss detection model is continuously trained.
In one embodiment, fig. 2 is a flowchart of a wear detection model training method according to another embodiment, and as shown in fig. 2, a process of labeling a wear category for an appearance image of the smart device in step S101 to establish an appearance image data set includes steps S200 and S201:
s200, selecting an intelligent equipment appearance image of a data set to be established according to the proportion of the average loss categories;
the intelligent device appearance images are classified according to the loss categories, an appearance image data set is established, intelligent device appearance images of all the loss categories are included in the appearance image data set, and the intelligent device appearance images of all the loss categories are the same in number. Use the loss classification including the mar, fall lacquer and explode as an example, the smart machine outward appearance quantity of mar: the number of appearances of the intelligent paint falling device is as follows: the number of the appearance of the intelligent equipment which is exposed to explosion is 1:1: 1.
S201, determining the proportion of a training set and a test set in the appearance image of the intelligent device to be established to determine the appearance image data set.
The setting of the training set and the test set is adaptive to the deep neural network of the subsequent steps, and the proportion of the training set and the test set is also adaptive to the deep neural network.
In one embodiment, the ratio of training set to test set is 9: 1.
And S102, establishing a deep neural network by using the appearance image data set so as to train a loss detection model for detecting appearance loss of the intelligent equipment.
The appearance image data set comprises a training set and a testing set and the like, wherein the training set is provided with loss class labels. And training out a deep neural network based on the training.
In one embodiment, the deep neural network is a yolo (young Only Look one) neural network, including a YOLOV3 neural network or a YOLOV5 neural network. As a preferred embodiment, the deep neural network is a Gaussian (Gaussian) based YOLOV3 neural network.
The Yolov3 neural network algorithm is improved by utilizing the normal distribution characteristic of Gaussian distribution, so that the network can output the uncertainty of each detection box, and the accuracy of the network is improved. Based on this deep neural network, in one embodiment, fig. 3 is a flowchart of a loss detection model training method according to yet another embodiment, and as shown in fig. 3, a process of building a deep neural network using the appearance image dataset in step S102 to train a loss detection model for detecting a loss category of an appearance of a smart device includes steps S300 to S303:
s300, determining the number of filters of the deep neural network according to the number of the loss categories;
s301, modifying the size of the image in the appearance image data set to be a multiple of the down-sampling frequency power of the step length;
s302, modifying impulse size, pooling algorithm, resetting loss function and learning rate;
and S303, adding a prediction layer for detecting the appearance loss to determine a loss detection model with optimal weight.
For convenience of explanation of the present embodiment, a specific application example is explained below — in the loss Detection model training method of another embodiment, based on the gaussian _ YOLOV3 neural network, when training is calculated according to the number C of loss classes, the number 36((1 frame confidence object +8 position outputs + C classes) x3 priori frames bbox) of filters of a cfg script file modifies the image size to be a multiple of 32 (5 times down-sampling the image in the YOLOV3 network structure, each time the sampling step is 2, so the input image size is modified to be a multiple of 32 (5 times of 2)), modifies the impulse momentum size to change a model optimizer, resets a loss function, modifies a pooled operation algorithm, resets a learning rate LR, and adds a prediction layer Detection to detect loss of a back image of a mobile phone, thereby obtaining a Detection model with optimal weight.
In the loss detection model training method in any embodiment, after the appearance image of the smart device is obtained, the loss category is labeled for the appearance image of the smart device, so as to establish an appearance image data set. Further, a deep neural network is built using the appearance image dataset to train a loss detection model that can be used to detect appearance loss of the smart device. Based on the method, a loss detection model is trained in advance through a large number of intelligent device appearance images serving as experience data, when the loss degree requirement for detecting the intelligent device to be detected is generated, the loss degree of the intelligent device to be detected can be rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent device recovery.
The embodiment of the invention also provides a loss detection model training device.
Fig. 4 is a block diagram of a wear detection model training apparatus according to an embodiment, and as shown in fig. 4, the wear detection model training apparatus according to the embodiment includes a block 100, a block 101, and a block 102:
the image acquisition module 100 is used for acquiring an appearance image of the intelligent device;
a data set establishing module 101, configured to label a loss category for the smart device appearance image to establish an appearance image data set;
and the model training module 102 is used for establishing a deep neural network by using the appearance image data set so as to train a loss detection model which can be used for detecting the appearance loss of the intelligent device.
The loss detection model training device in any embodiment above, after obtaining the appearance image of the smart device, labels the loss category for the appearance image of the smart device to establish an appearance image dataset. Further, a deep neural network is built using the appearance image dataset to train a loss detection model that can be used to detect appearance loss of the smart device. Based on the method, a loss detection model is trained in advance through a large number of intelligent device appearance images serving as experience data, when the loss degree requirement for detecting the intelligent device to be detected is generated, the loss degree of the intelligent device to be detected can be rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent device recovery.
Embodiments of the present invention further provide a computer storage medium, on which computer instructions are stored, and when the instructions are executed by a processor, the method for training a wear detection model according to any of the above embodiments is implemented.
Those skilled in the art will understand that: all or part of the steps for implementing the method embodiments may be implemented by hardware related to program instructions, and the program may be stored in a computer readable storage medium, and when executed, the program performs the steps including the method embodiments; and the aforementioned storage medium includes: various media that can store program codes, such as a removable Memory device, a Random Access Memory (RAM), a Read-Only Memory (ROM), a magnetic disk, and an optical disk.
Alternatively, the integrated unit of the present invention may be stored in a computer-readable storage medium if it is implemented in the form of a software functional module and sold or used as a separate product. Based on such understanding, the technical solutions of the embodiments of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a terminal, or a network device) to execute all or part of the methods of the embodiments of the present invention. And the aforementioned storage medium includes: a removable storage device, a RAM, a ROM, a magnetic or optical disk, or various other media that can store program code.
Corresponding to the computer storage medium, in an embodiment, there is provided another computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any one of the wear detection model training methods in the embodiments.
After the appearance image of the intelligent device is obtained, the computer device marks the loss category for the appearance image of the intelligent device so as to establish an appearance image data set. Further, a deep neural network is built using the appearance image dataset to train a loss detection model that can be used to detect appearance loss of the smart device. Based on the method, a loss detection model is trained in advance through a large number of intelligent device appearance images serving as experience data, when the loss degree requirement for detecting the intelligent device to be detected is generated, the loss degree of the intelligent device to be detected can be rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent device recovery.
The embodiment of the invention also provides a loss detection method based on the loss detection model of any one of the embodiments.
Fig. 5 is a flowchart illustrating a loss detection method according to an embodiment, and as shown in fig. 5, the loss detection method according to an embodiment includes steps S400 to S402:
s400, obtaining an appearance image to be detected; the appearance image to be detected is an appearance image of the intelligent equipment to be detected;
s401, conducting loss category probability prediction on the appearance image to be detected according to a loss detection model;
s402, determining the appearance loss degree of the intelligent device to be tested according to the loss category probability.
The intelligent device to be detected is the intelligent device which needs to be detected in the recovery process. And inputting the appearance image to be tested into the trained loss detection model in any one of the embodiments for model calculation.
The loss category probability prediction is given in a weight form or a priority form, so that the appearance loss degree can be intuitively known or digitalized.
Taking a loss detection model based on a Gaussia _ YOLOV3 neural network as an example, fig. 6 is a convolution flow chart of an appearance image to be detected in the loss detection model, and as shown in fig. 6, the appearance image to be detected is subjected to a series of convolution (Conv) operations after being Input into the neural network structure of the loss detection model. In three prediction layer Detection feature grids with different sizes, namely 13x13, 26x26 and 52x52, namely loss category prediction is carried out on an appearance image to be measured in each frame (tx, ty, tw and th), wherein tx and ty are predicted coordinate offsets, and tw and th are image scale scaling. And then, reducing the number of the prediction frames according to the target confidence coefficient threshold, non-maximum value inhibition, loss functions and other steps, classifying each frame into an array, and finally, calculating the length and the probability of each target loss category (scratch, paint removal and explosion), namely the array consisting of the frames, to obtain the probability of each of the three loss categories, wherein the loss category with the probability greater than 0 is the result of all the loss categories detected by the appearance image to be detected.
According to the loss detection method, after the appearance image to be detected is obtained, loss class probability prediction is carried out on the appearance image to be detected according to the loss detection model, and the appearance loss degree of the intelligent device to be detected is determined according to the loss class probability. Based on this, the loss degree of the intelligent equipment to be tested is rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent equipment recovery.
The embodiment of the invention also provides a loss detection device.
Fig. 7 is a block configuration diagram of a loss detection apparatus according to an embodiment, and as shown in fig. 7, the loss detection apparatus according to the embodiment includes a block 200, a block 201, and a block 202:
the to-be-detected image acquisition module 200 is used for acquiring an appearance image to be detected; the appearance image to be detected is an appearance image of the intelligent equipment to be detected;
a probability prediction module 201, configured to perform loss class probability prediction on the to-be-detected appearance image according to a loss detection model;
and the loss determining module 202 is configured to determine a loss degree of the to-be-tested smart device according to the loss category probability.
After the appearance image to be detected is obtained, the loss detection device predicts the loss class probability of the appearance image to be detected according to the loss detection model, and determines the appearance loss degree of the intelligent device to be detected according to the loss class probability. Based on this, the loss degree of the intelligent equipment to be tested is rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent equipment recovery.
The embodiment of the present invention further provides another computer storage medium, on which computer instructions are stored, and the instructions, when executed by a processor, implement the wear detection method of any one of the above embodiments.
Those skilled in the art will understand that: all or part of the steps for implementing the method embodiments may be implemented by hardware related to program instructions, and the program may be stored in a computer readable storage medium, and when executed, the program performs the steps including the method embodiments; and the aforementioned storage medium includes: various media that can store program codes, such as a removable Memory device, a Random Access Memory (RAM), a Read-Only Memory (ROM), a magnetic disk, and an optical disk.
Alternatively, the integrated unit of the present invention may be stored in a computer-readable storage medium if it is implemented in the form of a software functional module and sold or used as a separate product. Based on such understanding, the technical solutions of the embodiments of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a terminal, or a network device) to execute all or part of the methods of the embodiments of the present invention. And the aforementioned storage medium includes: a removable storage device, a RAM, a ROM, a magnetic or optical disk, or various other media that can store program code.
Corresponding to the computer storage medium, in an embodiment, there is provided another computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any one of the wear detection methods in the embodiments.
After the to-be-detected appearance image is obtained, the computer equipment carries out loss category probability prediction on the to-be-detected appearance image according to the loss detection model, and determines the appearance loss degree of the to-be-detected intelligent equipment according to the loss category probability. Based on this, the loss degree of the intelligent equipment to be tested is rapidly determined through the loss detection model trained in advance, and loss degree reference is stably and efficiently provided for intelligent equipment recovery.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above examples only show some embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A loss detection model training method is characterized by comprising the following steps:
acquiring an appearance image of the intelligent device;
marking a loss category for the appearance image of the intelligent equipment to establish an appearance image data set;
and establishing a deep neural network by using the appearance image data set so as to train a loss detection model which can be used for detecting the appearance loss of the intelligent equipment.
2. The loss detection model training method according to claim 1, wherein the process of labeling the smart device appearance image with a loss category to establish an appearance image dataset comprises the steps of:
selecting an intelligent equipment appearance image of a data set to be established according to the proportion of the average loss categories;
and determining the proportion of a training set and a test set in the appearance image of the intelligent equipment with the data set to be established so as to determine the appearance image data set.
3. The wear detection model training method of claim 2, wherein the wear categories include scratch, paint drop, and pop.
4. The loss detection model training method of claim 2, wherein the ratio of the training set to the test set is 9: 1.
5. The loss detection model training method of any one of claims 1 to 4, wherein the deep neural network comprises a YOLO neural network.
6. The loss detection model training method according to claim 5, wherein the process of building a deep neural network using the appearance image dataset to train a loss detection model for detecting loss classes of an appearance of a smart device comprises the steps of:
determining the number of filters of the deep neural network according to the number of the loss categories;
modifying the image size in the appearance image data set to be a multiple of the step length to the power of the down-sampling times;
modifying impulse size, pooling algorithm, resetting loss function and learning rate;
and adding a prediction layer for detecting the appearance loss to determine a loss detection model with optimal weight.
7. The loss detection model training method of claim 5, wherein the YOLO neural network comprises a YOLOV3 neural network or a YOLOV5 neural network.
8. A loss detection model training device, comprising:
the image acquisition module is used for acquiring an appearance image of the intelligent equipment;
the data set establishing module is used for marking the loss category for the appearance image of the intelligent equipment so as to establish an appearance image data set;
and the model training module is used for establishing a deep neural network by using the appearance image data set so as to train a loss detection model which can be used for detecting the appearance loss of the intelligent equipment.
9. A method of loss detection, comprising the steps of:
acquiring an appearance image to be detected; the appearance image to be detected is an appearance image of the intelligent equipment to be detected;
performing loss class probability prediction on the appearance image to be detected according to a loss detection model;
and determining the appearance loss degree of the intelligent device to be tested according to the loss category probability.
10. A wear detection apparatus, comprising:
the to-be-detected image acquisition module is used for acquiring an appearance image to be detected; the appearance image to be detected is an appearance image of the intelligent equipment to be detected;
the probability prediction module is used for performing loss category probability prediction on the appearance image to be detected according to the loss detection model;
and the loss determining module is used for determining the loss degree of the intelligent equipment to be tested according to the loss category probability.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011586533.4A CN112634245A (en) | 2020-12-28 | 2020-12-28 | Loss detection model training method, loss detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011586533.4A CN112634245A (en) | 2020-12-28 | 2020-12-28 | Loss detection model training method, loss detection method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112634245A true CN112634245A (en) | 2021-04-09 |
Family
ID=75286021
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011586533.4A Pending CN112634245A (en) | 2020-12-28 | 2020-12-28 | Loss detection model training method, loss detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112634245A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113298078A (en) * | 2021-06-22 | 2021-08-24 | 广州绿怡信息科技有限公司 | Equipment screen fragmentation detection model training method and equipment screen fragmentation detection method |
| CN113591963A (en) * | 2021-07-23 | 2021-11-02 | 广州绿怡信息科技有限公司 | Equipment side loss detection model training method and equipment side loss detection method |
| US11922467B2 (en) | 2020-08-17 | 2024-03-05 | ecoATM, Inc. | Evaluating an electronic device using optical character recognition |
| US12033454B2 (en) | 2020-08-17 | 2024-07-09 | Ecoatm, Llc | Kiosk for evaluating and purchasing used electronic devices |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180174046A1 (en) * | 2016-12-15 | 2018-06-21 | Beijing Kuangshi Technology Co., Ltd. | Target detection method and device, neural network training method and device |
| CN108921811A (en) * | 2018-04-03 | 2018-11-30 | 阿里巴巴集团控股有限公司 | Detect method and apparatus, the article damage detector of article damage |
| CN110059734A (en) * | 2019-04-02 | 2019-07-26 | 唯思科技（北京）有限公司 | A kind of training method, object identification method, device, robot and the medium of target identification disaggregated model |
| CN111275660A (en) * | 2018-12-05 | 2020-06-12 | 合肥欣奕华智能机器有限公司 | Defect detection method and device for flat panel display |
| CN111294454A (en) * | 2020-03-23 | 2020-06-16 | 广州绿怡信息科技有限公司 | Mobile phone loss detection device and method |
| WO2020155518A1 (en) * | 2019-02-03 | 2020-08-06 | 平安科技（深圳）有限公司 | Object detection method and device, computer device and storage medium |
| CN111612763A (en) * | 2020-05-20 | 2020-09-01 | 重庆邮电大学 | Mobile phone screen defect detection method, device and system, computer equipment and medium |
| CN111898651A (en) * | 2020-07-10 | 2020-11-06 | 江苏科技大学 | Tree detection method based on Tiny Yolov3 algorithm |
-
2020
- 2020-12-28 CN CN202011586533.4A patent/CN112634245A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180174046A1 (en) * | 2016-12-15 | 2018-06-21 | Beijing Kuangshi Technology Co., Ltd. | Target detection method and device, neural network training method and device |
| CN108921811A (en) * | 2018-04-03 | 2018-11-30 | 阿里巴巴集团控股有限公司 | Detect method and apparatus, the article damage detector of article damage |
| CN111275660A (en) * | 2018-12-05 | 2020-06-12 | 合肥欣奕华智能机器有限公司 | Defect detection method and device for flat panel display |
| WO2020155518A1 (en) * | 2019-02-03 | 2020-08-06 | 平安科技（深圳）有限公司 | Object detection method and device, computer device and storage medium |
| CN110059734A (en) * | 2019-04-02 | 2019-07-26 | 唯思科技（北京）有限公司 | A kind of training method, object identification method, device, robot and the medium of target identification disaggregated model |
| CN111294454A (en) * | 2020-03-23 | 2020-06-16 | 广州绿怡信息科技有限公司 | Mobile phone loss detection device and method |
| CN111612763A (en) * | 2020-05-20 | 2020-09-01 | 重庆邮电大学 | Mobile phone screen defect detection method, device and system, computer equipment and medium |
| CN111898651A (en) * | 2020-07-10 | 2020-11-06 | 江苏科技大学 | Tree detection method based on Tiny Yolov3 algorithm |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11922467B2 (en) | 2020-08-17 | 2024-03-05 | ecoATM, Inc. | Evaluating an electronic device using optical character recognition |
| US12033454B2 (en) | 2020-08-17 | 2024-07-09 | Ecoatm, Llc | Kiosk for evaluating and purchasing used electronic devices |
| CN113298078A (en) * | 2021-06-22 | 2021-08-24 | 广州绿怡信息科技有限公司 | Equipment screen fragmentation detection model training method and equipment screen fragmentation detection method |
| CN113591963A (en) * | 2021-07-23 | 2021-11-02 | 广州绿怡信息科技有限公司 | Equipment side loss detection model training method and equipment side loss detection method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112634245A (en) | Loss detection model training method, loss detection method and device | |
| CN108596221B (en) | Image recognition method and device for scale reading | |
| CN109901996B (en) | Auxiliary test method and device, electronic equipment and readable storage medium | |
| CN111524112B (en) | Steel chasing identification method, system, equipment and medium | |
| CN106846362A (en) | A kind of target detection tracking method and device | |
| CN111582358B (en) | Training method and device for house type recognition model, and house type weight judging method and device | |
| CN115713715B (en) | Human behavior recognition method and recognition system based on deep learning | |
| CN112614117A (en) | Equipment region extraction model training method, equipment region extraction method and device | |
| CN110717407B (en) | Face recognition method, device and storage medium based on lip language password | |
| CN116952958B (en) | Defect detection method, device, electronic equipment and storage medium | |
| CN112308069A (en) | Click test method, device, equipment and storage medium for software interface | |
| CN112990318A (en) | Continuous learning method, device, terminal and storage medium | |
| CN112348808A (en) | Screen perspective detection method and device | |
| CN112801489B (en) | Litigation case risk detection method, litigation case risk detection device, litigation case risk detection equipment and readable storage medium | |
| CN114443483A (en) | Test method and device of artificial intelligence system, electronic equipment and medium | |
| CN117726862A (en) | Model training method, device and storage medium applied to industrial detection | |
| CN113610832B (en) | Logo defect detection method, device, equipment and storage medium | |
| CN114841255A (en) | Detection model training method, device, equipment, storage medium and program product | |
| CN104680123A (en) | Object identification device, object identification method and program | |
| CN115546108A (en) | Intelligent detection method for appearance quality of automobile tire based on edge cloud cooperation and AR | |
| CN109583590B (en) | Data processing method and data processing device | |
| CN113901996A (en) | Equipment screen perspective detection model training method and equipment screen perspective detection method | |
| CN112819079A (en) | Model sampling algorithm matching method and device and electronic equipment | |
| CN113537342B (en) | Method and device for detecting object in image, storage medium and terminal | |
| CN108696722A (en) | A kind of target monitoring method, system and 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 |