CN112016630A - Training method, device and equipment based on image classification model and storage medium - Google Patents
Training method, device and equipment based on image classification model and storage medium Download PDFInfo
- Publication number
- CN112016630A CN112016630A CN202010925560.3A CN202010925560A CN112016630A CN 112016630 A CN112016630 A CN 112016630A CN 202010925560 A CN202010925560 A CN 202010925560A CN 112016630 A CN112016630 A CN 112016630A
- Authority
- CN
- China
- Prior art keywords
- sample
- image
- classification model
- training
- image classification
- 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.)
- Granted
Links
- 238000012549 training Methods 0.000 title claims abstract description 218
- 238000013145 classification model Methods 0.000 title claims abstract description 184
- 238000000034 method Methods 0.000 title claims abstract description 55
- 230000015572 biosynthetic process Effects 0.000 claims abstract description 114
- 238000003786 synthesis reaction Methods 0.000 claims abstract description 114
- 238000012545 processing Methods 0.000 claims description 67
- 230000006870 function Effects 0.000 claims description 43
- 238000004422 calculation algorithm Methods 0.000 claims description 41
- 230000003416 augmentation Effects 0.000 claims description 26
- 238000003708 edge detection Methods 0.000 claims description 15
- 238000013528 artificial neural network Methods 0.000 claims description 13
- 238000013473 artificial intelligence Methods 0.000 abstract description 7
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 abstract description 5
- 238000005516 engineering process Methods 0.000 abstract description 5
- 230000009466 transformation Effects 0.000 abstract description 3
- 230000003321 amplification Effects 0.000 abstract 2
- 238000003199 nucleic acid amplification method Methods 0.000 abstract 2
- 230000002194 synthesizing effect Effects 0.000 abstract 1
- 230000008569 process Effects 0.000 description 11
- 238000004891 communication Methods 0.000 description 5
- 238000002372 labelling Methods 0.000 description 5
- 238000007635 classification algorithm Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 239000007787 solid Substances 0.000 description 4
- 230000001149 cognitive effect Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 230000003190 augmentative effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of artificial intelligence, and discloses a training method, a device, equipment and a storage medium based on an image classification model, wherein the method comprises the steps of respectively obtaining a background image sample, a foreground image sample and an application image sample of an object to be classified, and constructing a function synthesis sample according to the application image sample; fitting the background image sample and the foreground image sample to obtain a synthetic training sample; and training the image classification model to be trained by synthesizing the training sample and the function synthesis sample to obtain a target image classification model. Compared with the existing image classification model obtained by amplifying the training sample without amplification or simple affine transformation, the method has the advantages that the background image sample and the foreground image sample are processed respectively and then are fitted, so that the sample amplification number is increased exponentially, and the classification precision of the image classification model is greatly improved. In addition, the invention also relates to a block chain technology, and the background image samples and the foreground image samples can be stored in the block chain.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a training method, a training device, training equipment and a storage medium based on an image classification model.
Background
Image classification is an important part in the field of artificial intelligence, most application systems based on machine vision need an image classification algorithm to judge the attributes of images collected by a camera and then enter other processing flows, so that the image classification capability is the basic capability in the field of machine vision.
Part of the existing image classification is based on a cognitive manual design classification algorithm of a human engineer, the algorithm debugging process is complex, professional talents are needed, and time and labor are consumed. The other part is that a neural network is used for training an image classification model, certain manpower is needed for collecting and marking samples, when the samples are difficult to collect and the sample amount is insufficient, the samples are not amplified or are amplified only based on simple affine transformation, and the accuracy of the obtained model is poor. If the background in the scene needing to be identified is single, but the background of the training sample is too variable, too much background information can be contained in the actual model, and the object information needing to be classified is insufficient, so that more misclassifications are caused in actual application, and the model precision is improved only by improving the method for simply increasing the sample by 10 times of the increase of the sample amount, so that the sample acquisition and labeling cost is too high, and the whole working efficiency of the algorithm improvement is difficult to improve.
The above is only for the purpose of assisting understanding of the technical aspects of the present invention, and does not represent an admission that the above is prior art.
Disclosure of Invention
The invention mainly aims to provide a training method, a training device, equipment and a storage medium based on an image classification model, and aims to solve the technical problems that in the prior art, the accuracy of the image classification model is poor and the efficiency of the improvement work of the image classification model is difficult to improve due to the fact that training samples are difficult to acquire or are too variable.
In order to achieve the above object, the present invention provides a training method based on an image classification model, which comprises the following steps:
respectively acquiring a background image sample and a foreground image sample of an object to be classified in a preset background environment;
acquiring an application image sample of the object to be classified in a preset application environment, and constructing a function synthesis sample according to the application image sample;
fitting according to the background image sample and the foreground image sample to obtain a synthetic training sample;
constructing an image classification model to be trained of the object to be classified based on a preset neural network;
and training the image classification model to be trained through the synthetic training sample and the function synthetic sample to obtain a target image classification model.
Preferably, the step of respectively obtaining a background image sample and a foreground image sample of the object to be classified in a preset background environment includes:
determining the image acquisition quantity according to the diversity condition of a preset background environment;
acquiring background image samples of the object to be classified in the preset background environment according to the image acquisition quantity;
and determining a foreground environment according to the preset background environment, and acquiring a foreground image sample in the foreground environment.
Preferably, the step of obtaining an application image sample of the object to be classified in a preset application environment and constructing a function synthesis sample according to the application image sample includes:
determining a preset classification number of the objects to be classified in a preset application environment;
acquiring an application image sample of the object to be classified in the preset application environment according to the preset classification quantity;
and constructing a function synthesis sample according to the distance information between the application image samples.
Preferably, the step of obtaining a synthetic training sample by fitting the background image sample and the foreground image sample includes:
segmenting the background image sample by a preset edge detection algorithm to obtain a target foreground image;
classifying and combining the background image sample and the target foreground image according to a preset augmentation algorithm to obtain an initial synthesis training sample;
and performing edge processing on the initial synthesis training sample according to the foreground image sample to obtain a synthesis training sample.
Preferably, the step of performing edge processing on the initial synthesis training sample according to the foreground image sample to obtain a synthesis training sample includes:
covering the foreground image sample on the background image sample, and acquiring the relative position of the foreground image sample relative to the background image sample;
obtaining a synthetic edge area of an initial synthetic training sample according to the edge area of the foreground image sample and the relative position;
performing transparency processing on the synthetic edge area to obtain an initial synthetic training sample after transparency processing;
and taking the initial synthetic training sample after transparency processing as a synthetic training sample.
Preferably, the step of performing transparency processing on the synthesized edge region to obtain an initial synthesized training sample after transparency processing includes:
acquiring gray values of all pixel points of the background image sample in the synthetic edge region;
acquiring gray values of all pixel points of the foreground image sample in the synthetic edge region;
performing transparency processing on the gray value of each pixel point of the background image sample and the gray value of each pixel point of the foreground image sample based on preset transparency to obtain a pixel synthesis value of each pixel point in the synthesis edge area;
and constructing an initial synthesis training sample after transparency processing according to the pixel synthesis value.
Preferably, the step of training the image classification model to be trained by the synthetic training sample and the functional synthetic sample to obtain a target image classification model includes:
updating network parameters of an image classification model to be trained through the synthetic training sample to obtain a basic image classification model;
obtaining classifier parameters for image classification, updating the classifier parameters of the basic image classification model according to the classifier parameters, and obtaining an updated basic image classification model;
and updating the network parameters of the updated basic image classification model through the function synthesis sample to obtain a target image classification model.
In addition, in order to achieve the above object, the present invention further provides an image classification model-based training apparatus, comprising:
the acquisition module is used for respectively acquiring a background image sample and a foreground image sample of an object to be classified in a preset background environment;
the acquisition module is further used for acquiring an application image sample of the object to be classified in a preset application environment and constructing a function synthesis sample according to the application image sample;
the fitting module is used for fitting according to the background image sample and the foreground image sample to obtain a synthetic training sample;
the building module is used for building an image classification model to be trained of the object to be classified based on a preset neural network;
and the training module is used for training the image classification model to be trained through the synthetic training sample and the function synthetic sample to obtain a target image classification model.
In addition, to achieve the above object, the present invention further provides an image classification model-based training apparatus, including: a memory, a processor, and an image classification model-based training program stored on the memory and executable on the processor, the image classification model-based training program configured to implement the steps of the image classification model-based training method as described above.
In addition, to achieve the above object, the present invention further provides a storage medium, on which an image classification model based training program is stored, and the image classification model based training program, when executed by a processor, implements the steps of the image classification model based training method as described above.
The method comprises the steps of respectively obtaining a background image sample, a foreground image sample and an application image sample of an object to be classified, constructing a function synthesis sample according to the application image sample, fitting the background image sample and the foreground image sample, and obtaining a synthesis training sample; constructing an image classification model to be trained of an object to be classified; and training the image classification model to be trained through the fitted synthetic training sample and the function synthetic sample to obtain a target image classification model. Compared with the prior art that the training samples are not augmented or are augmented only based on simple affine transformation, the accuracy of the obtained image classification model is poor, and the sample augmentation mode that the background image samples and the foreground image samples are respectively processed and then fitted is adopted, so that the number of sample augmentation can be exponentially increased, the classification accuracy of the image classification model is greatly improved, and the technical problems that the accuracy of the image classification model is poor and the efficiency of the image classification model improvement work is difficult to improve due to the fact that the training samples are difficult to collect or too variable in the prior art are solved.
Drawings
FIG. 1 is a schematic diagram of an apparatus in a hardware operating environment according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating a first embodiment of the training method based on an image classification model according to the present invention;
FIG. 3 is a flowchart illustrating a second embodiment of the training method based on image classification model according to the present invention;
FIG. 4 is a flowchart illustrating a third embodiment of the training method based on image classification models according to the present invention;
FIG. 5 is a block diagram of a training apparatus based on an image classification model according to a first embodiment of the present invention.
The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Referring to fig. 1, fig. 1 is a schematic device structure diagram of a hardware operating environment according to an embodiment of the present invention.
As shown in fig. 1, the apparatus may include: a processor 1001, such as a Central Processing Unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Wherein a communication bus 1002 is used to enable connective communication between these components. The user interface 1003 may include a Display screen (Display), an input unit such as a Keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface, a wireless interface. The network interface 1004 may optionally include a standard wired interface, a WIreless interface (e.g., a WIreless-FIdelity (WI-FI) interface). The Memory 1005 may be a Random Access Memory (RAM) Memory, or may be a Non-Volatile Memory (NVM), such as a disk Memory. The memory 1005 may alternatively be a storage device separate from the processor 1001.
Those skilled in the art will appreciate that the configuration shown in fig. 1 does not constitute a limitation of the device and may include more or fewer components than those shown, or some components may be combined, or a different arrangement of components.
As shown in fig. 1, a memory 1005, which is a storage medium, may include therein an operating system, a data storage module, a network communication module, a user interface module, and a training program based on an image classification model.
In the device shown in fig. 1, the network interface 1004 is mainly used for data communication with a network server; the user interface 1003 is mainly used for data interaction with a user; the processor 1001 and the memory 1005 of the apparatus of the present invention may be provided in the apparatus, and the apparatus calls the training program based on the image classification model stored in the memory 1005 through the processor 1001 and executes the training method based on the image classification model provided by the embodiment of the present invention.
An embodiment of the present invention provides a training method based on an image classification model, and referring to fig. 2, fig. 2 is a schematic flow chart of a first embodiment of the training method based on the image classification model according to the present invention.
In this embodiment, the training method based on the image classification model includes the following steps:
step S10: respectively obtaining a background image sample and a foreground image sample of an object to be classified in a preset background environment.
It should be noted that the execution subject of the method of the present embodiment may be a training device based on an image classification model. Respectively acquiring a background image sample and a foreground image sample of an object to be classified in a preset background environment, wherein the preset background environment can be the environment where the object to be classified is located, and the image acquisition quantity can be determined according to the diversity condition of the preset background environment; acquiring background image samples of the object to be classified in the preset background environment according to the image acquisition quantity; and determining a foreground environment according to the preset background environment, and acquiring a foreground image sample in the foreground environment.
Specifically, background pictures are acquired in the environment where the object to be classified is located, and the number of image acquisition, that is, the number of background image samples to be acquired, is determined according to the change of conditions such as the lighting condition, the shooting angle, the shooting distance and the like of the preset background environment, wherein the image acquisition number is preferably in a range that completely covers the preset background environment.
It should be understood that the foreground environment is determined according to the preset background environment, the foreground environment may be a background environment as simple as possible, a foreground image sample of the object to be classified is collected in the background environment as simple as possible, as many foreground image samples of the object to be classified as possible are collected according to possible presentation ranges of the object to be classified in an application scene, such as different shooting distances and shooting angles, and image classification labeling is performed on the foreground image sample according to different objects to be classified, so that training of an image classification model is facilitated.
It should be emphasized that, in order to further ensure the privacy and security of the background image samples and the foreground image samples, the background image samples and the foreground image samples may also be stored in nodes of a block chain.
Step S20: and acquiring an application image sample of the object to be classified in a preset application environment, and constructing a function synthesis sample according to the application image sample.
It is easy to understand that, an application image sample of the object to be classified is obtained in a preset application environment, wherein the preset application environment can be the application environment of the object to be classified, and the preset classification number of the object to be classified is determined in the preset application environment; acquiring an application image sample of the object to be classified in the preset application environment according to the preset classification quantity; and constructing a function synthesis sample according to the distance information between the application image samples. The preset number of classifications is to take as many application image samples as the conditions allow.
Specifically, objects to be classified are shot in an application environment, each object in the objects to be classified needs to be shot and subjected to image classification marking, as many application image samples as possible are shot under conditions, function synthesis samples are constructed according to distance information between the application image samples, the function synthesis samples are used for training and verifying of a subsequent image classification model, and the more application image samples actually shot, the higher the accuracy of the image classification model obtained through training is.
Step S30: and fitting according to the background image sample and the foreground image sample to obtain a synthetic training sample.
It should be noted that, fitting is performed according to the shot background image sample and the shot foreground image sample, so as to obtain a synthetic training sample. One feasible fitting method is to use a foreground image sample to cover a background image sample and perform fitting in a manner of performing background transparency processing near the edge to obtain a synthetic training sample, and the specific process may be as follows: segmenting the background image sample by a preset edge detection algorithm to obtain a target foreground image; classifying and combining the background image sample and the target foreground image according to a preset augmentation algorithm to obtain an initial synthesis training sample; and performing edge processing on the initial synthesis training sample according to the foreground image sample to obtain a synthesis training sample.
Specifically, a foreground environment is determined according to the preset background environment, the foreground environment can be a background environment as simple as possible, a foreground image sample of an object to be classified is collected in the background environment as simple as possible, the shot foreground image sample uses the background environment as simple as possible, a target foreground image can be obtained through a matting technology, for example, the background image sample can be segmented through a preset edge detection algorithm to obtain a target foreground image, namely, a foreground and a solid background of the background image sample are segmented through a simple edge detection algorithm to obtain the target foreground image.
Further, classifying and combining the background image sample and the target foreground image according to a preset augmentation algorithm to obtain an initial synthesis training sample; the preset augmentation algorithm may include such augmentation algorithms as blurring, rotation, distortion and color cast, and performs random combination after the background image sample and the target foreground image are classified, so as to obtain an initial synthesized training sample, make as many target foreground images as possible appear in as many background image sample combinations as possible, generally directly cover the target foreground images in partial areas of the background image sample, and perform classification labeling on the synthesized initial synthesized training sample according to different target foreground image classifications.
Further, edge processing is performed on the synthesized initial synthesis training sample, and generally, an object edge area of the synthesized initial synthesis training sample can be obtained according to the edge of the foreground image sample and the relative position on the background image sample, and transparency processing is performed on the object edge area to obtain the synthesis training sample.
Step S40: and constructing an image classification model to be trained of the object to be classified based on a preset neural network.
It should be understood that the image classification model to be trained of the object to be classified is constructed based on a preset neural network, or the image classification model to be trained of the object to be classified is constructed according to an actual requirement model, the image classification model to be trained of the object to be classified may be constructed based on a classification algorithm of cognitive artificial design of a human engineer, the image classification model to be trained of the object to be classified may also be an image classification model trained based on a neural network, the synthesized training sample and the function synthesized sample in this embodiment may train the two types of image classification models to be trained, obtain a target image classification model, improve the precision of the image classification model, and the image classification model to be trained of the object to be classified may also be an image classification model to be trained obtained in other manners, which is not limited in this embodiment.
Step S50: and training the image classification model to be trained through the synthetic training sample and the function synthetic sample to obtain a target image classification model.
It is easy to understand that, the image classification model to be trained is trained through the synthesis training sample and the function synthesis sample to obtain a target image classification model, and the training of the image classification model to be trained can adopt a variety of ways, and this embodiment is described by the following training ways: updating network parameters of an image classification model to be trained through the synthetic training sample to obtain a basic image classification model; obtaining classifier parameters for image classification, updating the classifier parameters of the basic image classification model according to the classifier parameters, and obtaining an updated basic image classification model; and updating the network parameters of the updated basic image classification model through the function synthesis sample to obtain a target image classification model.
It should be noted that, in this embodiment, other training manners may also be used to train the to-be-trained image classification model through a synthetic training sample and a function synthetic sample, so as to obtain the target image classification model, which is not limited in this embodiment. Image classification is an important part in the field of artificial intelligence, most application systems based on machine vision need an image classification model, attribute judgment is carried out on images collected by a camera, then other processing flows are carried out, and the classification capability of the image classification model is the basic capability in the field of machine vision. The embodiment enables the sample augmentation number to be increased exponentially, greatly improves the classification precision of the image classification model, and can be applied to the field of artificial intelligence of machine vision.
In the embodiment, a background image sample and a foreground image sample of an object to be classified are respectively obtained in a preset background environment; acquiring an application image sample of the object to be classified in a preset application environment, and constructing a function synthesis sample according to the application image sample; fitting according to the background image sample and the foreground image sample to obtain a synthetic training sample; constructing an image classification model to be trained of the object to be classified based on a preset neural network; and training the image classification model to be trained through the synthetic training sample and the function synthetic sample to obtain a target image classification model. This embodiment is compared in current training sample and is not enlarged or only enlarge based on simple affine transform, the image classification model precision that obtains is relatively poor, adopt background image sample and foreground image sample to handle the sample mode of enlarging of fitting again after handling respectively, can make sample number of enlarging obtain exponential increase, greatly promote image classification model's classification precision, it is poor to have solved the image classification model precision that prior art training sample is difficult to gather or too changeable leads to, the image classification model improves the technical problem that the efficiency of work is difficult to improve.
Referring to fig. 3, fig. 3 is a flowchart illustrating a training method based on an image classification model according to a second embodiment of the present invention.
Based on the first embodiment described above, in the present embodiment, the step S30 includes:
step S301: and segmenting the background image sample by a preset edge detection algorithm to obtain a target foreground image.
It should be noted that, fitting is performed according to the shot background image sample and the shot foreground image sample, so as to obtain a synthetic training sample. One possible fitting method is to use a foreground image sample to cover a background image sample and perform background transparency processing near the edge to obtain a synthetic training sample.
Specifically, the background image sample is segmented by a preset edge detection algorithm to obtain a target foreground image, and the preset edge detection algorithm may be a simple edge detection algorithm. The method comprises the steps of determining a foreground environment according to a preset background environment, wherein the foreground environment can be a background environment which is as simple as possible, collecting a foreground image sample of an object to be classified in the background environment which is as simple as possible, obtaining a target foreground image through a matting technology due to the fact that the shot foreground image sample uses the background environment which is as simple as possible, and obtaining the target foreground image by segmenting a foreground and a solid background of the background image sample through a simple edge detection algorithm. The foreground and the solid background of the background image sample may also be segmented by other algorithms to obtain the target foreground image, which is not limited in this embodiment.
Step S302: and carrying out classification combination on the background image sample and the target foreground image according to a preset augmentation algorithm to obtain an initial synthesis training sample.
It is easy to understand that the background image sample and the target foreground image are classified and combined according to a preset augmentation algorithm, where the preset augmentation algorithm may include any one of augmentation algorithms such as blur, rotation, distortion, and color cast, and the preset augmentation algorithm may include other augmentation algorithms, which is not limited in this embodiment, and the preset augmentation algorithm including blur, rotation, distortion, and color cast augmentation algorithm is described in this embodiment.
Specifically, the process of obtaining the initial synthesis training sample by classifying and combining the background image sample and the target foreground image according to the preset augmentation algorithm may be as follows: the background image sample and the target foreground image are classified and randomly combined according to fuzzy, rotation, distortion and color cast augmentation algorithms to obtain an initial synthesis training sample, so that the target foreground images are presented in the background image sample combination as much as possible, the target foreground images can be generally directly covered in partial areas of the background image sample, and the synthesized initial synthesis training sample is classified and labeled according to different target foreground image classifications.
Step S303: and performing edge processing on the initial synthesis training sample according to the foreground image sample to obtain a synthesis training sample.
It should be noted that, the edge processing is performed on the synthesized initial synthesis training sample, and generally, an object edge area of the synthesized initial synthesis training sample can be obtained according to the edge of the foreground image sample and the relative position on the background image sample, and the transparency processing is performed on the object edge area to obtain the synthesis training sample.
Specifically, the foreground image sample can be covered on the background image sample, and the relative position of the foreground image sample relative to the background image sample is obtained; obtaining a synthetic edge area of an initial synthetic training sample according to the edge area and the relative position of the foreground image sample; performing transparency processing on the synthetic edge area to obtain an initial synthetic training sample after transparency processing; and taking the initial synthesis training sample after transparency processing as a synthesis training sample.
In the embodiment, the background image sample is segmented through a preset edge detection algorithm to obtain a target foreground image; classifying and combining the background image sample and the target foreground image according to a preset augmentation algorithm to obtain an initial synthesis training sample; and performing edge processing on the initial synthesis training sample according to the foreground image sample to obtain a synthesis training sample. In the generation process of the synthetic training sample, the initial synthetic training sample can be subjected to edge processing according to the foreground image sample to obtain the synthetic training sample, because the transparency of the foreground image sample edge can have different value strategies, values can be taken randomly or uniformly in a certain step length in a certain range, and the like, when the to-be-trained image classification model is trained through the synthetic training sample and the function synthetic sample, the to-be-trained image classification model can further reduce the attention to the edge, so that more information such as the shape texture of the object can be obtained, and the classification precision of the trained target image classification model can be improved.
Referring to fig. 4, fig. 4 is a flowchart illustrating a training method based on an image classification model according to a third embodiment of the present invention.
Based on the second embodiment, in this embodiment, the step S303 includes:
step S3031: covering the foreground image sample on the background image sample, and acquiring the relative position of the foreground image sample relative to the background image sample.
It should be noted that, the edge processing is performed on the synthesized initial synthesis training sample, and generally, an object edge area of the synthesized initial synthesis training sample can be obtained according to the edge of the foreground image sample and the relative position on the background image sample, and the transparency processing is performed on the object edge area to obtain the synthesis training sample. And obtaining an object edge area of the initially synthesized training sample after synthesis, wherein the foreground image sample needs to be covered on the background image sample, and the relative position of the foreground image sample relative to the background image sample is obtained.
Step S3032: and obtaining a synthetic edge area of the initial synthetic training sample according to the edge area of the foreground image sample and the relative position.
It is easy to understand that, according to the edge area of the foreground image sample and the relative position of the foreground image sample with respect to the background image sample, the synthesized edge area of the initial synthesis training sample can be obtained, and after transparency processing is performed on the synthesized edge area, the initial synthesis training sample after transparency processing can be obtained. The initial synthesis training sample after transparency processing is used as a synthesis training sample, and when the image classification model to be trained is trained, the initial synthesis training sample after transparency processing can further enable the image classification model to be trained to reduce the attention to edges, and improve the classification precision of the trained target image classification model.
Step S3033: and performing transparency processing on the synthesized edge area to obtain an initial synthesis training sample after transparency processing.
It should be noted that, the process of performing transparency processing on the synthesized edge region to obtain an initial synthesized training sample after transparency processing may be as follows: acquiring gray values of all pixel points of the background image sample in the synthetic edge region; acquiring gray values of all pixel points of the foreground image sample in the synthetic edge region; performing transparency processing on the gray value of each pixel point of the background image sample and the gray value of each pixel point of the foreground image sample based on preset transparency to obtain a pixel synthesis value of each pixel point in the synthesis edge area; and constructing an initial synthesis training sample after transparency processing according to the pixel synthesis value.
Specifically, transparency processing is performed on the synthesized edge region, for example, in an RGB color space, when the transparency is alpha, the background image sample is B, the foreground image sample is a, and the synthesized initial synthesis training sample is C, in this embodiment, taking a green color G channel as an example, a pixel synthesis value of each pixel point in the synthesized edge region may be obtained, and a pixel synthesis value of a G pixel point in the synthesized edge region may be calculated by the following formula: g (1-alpha) × G (b) + alpha × G (a), where alpha is a preset transparency, G (a) is a gray value of a G pixel point of a foreground image sample, G (b) is a gray value of a G pixel point of a background image sample, and G (c) is a pixel synthesis value of a G pixel point in a synthesis edge region, and in an RGB color space, a pixel synthesis value of a red pixel point and a pixel synthesis value of a blue pixel point in the synthesis edge region may also be obtained according to the above manner, and an initial synthesis training sample after transparency processing may be constructed according to a pixel synthesis value of a green pixel point in the synthesis edge region, a pixel synthesis value of a red pixel point, and a pixel synthesis value of a blue pixel point. In addition, the transparency processing on the synthesized edge region may also be performed in an RGBW color space or other color spaces, which is not limited in this embodiment.
Step S3034: and taking the initial synthetic training sample after transparency processing as a synthetic training sample.
It should be understood that the initial synthesis training sample after transparency processing is used as a synthesis training sample, and since the transparency may have different value-taking strategies, values may be taken randomly within a certain range or uniformly at a certain step length, and the like, when the image classification model to be trained is trained through the synthesis training sample and the function synthesis sample, the attention to the edge of the image classification model to be trained may be further reduced, and information such as shape and texture of more objects may be obtained, so as to improve the classification accuracy of the trained target image classification model. In order to avoid too similar edge portions of the synthesized training samples, random preset transparency alpha values may be used within a certain range.
In this embodiment, the foreground image sample is covered on the background image sample, and the relative position of the foreground image sample with respect to the background image sample is obtained; obtaining a synthetic edge area of an initial synthetic training sample according to the edge area of the foreground image sample and the relative position; performing transparency processing on the synthetic edge area to obtain an initial synthetic training sample after transparency processing; and taking the initial synthetic training sample after transparency processing as a synthetic training sample. In the generation process of the synthetic training sample, transparency processing can be performed on the synthetic edge region to obtain an initial synthetic training sample after transparency processing, and different value-taking strategies can be provided for transparency, so that values can be taken randomly or uniformly in a certain step length in a certain range, and the like.
Furthermore, an embodiment of the present invention further provides a storage medium, where the storage medium stores an image classification model-based training program, and the image classification model-based training program, when executed by a processor, implements the steps of the image classification model-based training method as described above.
Referring to fig. 5, fig. 5 is a block diagram illustrating a first embodiment of an image classification model-based training apparatus according to the present invention.
As shown in fig. 5, the training apparatus based on the image classification model according to the embodiment of the present invention includes:
the acquiring module 10 is configured to acquire a background image sample and a foreground image sample of an object to be classified in a preset background environment, respectively.
It should be noted that the execution subject of the method of the present embodiment may be a training device based on an image classification model. Respectively acquiring a background image sample and a foreground image sample of an object to be classified in a preset background environment, wherein the preset background environment can be the environment where the object to be classified is located, and the image acquisition quantity can be determined according to the diversity condition of the preset background environment; acquiring background image samples of the object to be classified in the preset background environment according to the image acquisition quantity; and determining a foreground environment according to the preset background environment, and acquiring a foreground image sample in the foreground environment.
Specifically, background pictures are acquired in the environment where the object to be classified is located, and the number of image acquisition, that is, the number of background image samples to be acquired, is determined according to the change of conditions such as the lighting condition, the shooting angle, the shooting distance and the like of the preset background environment, wherein the image acquisition number is preferably in a range that completely covers the preset background environment.
It should be understood that the foreground environment is determined according to the preset background environment, the foreground environment may be a background environment as simple as possible, a foreground image sample of the object to be classified is collected in the background environment as simple as possible, as many foreground image samples of the object to be classified as possible are collected according to possible presentation ranges of the object to be classified in an application scene, such as different shooting distances and shooting angles, and image classification labeling is performed on the foreground image sample according to different objects to be classified, so that training of an image classification model is facilitated.
The obtaining module 10 is further configured to obtain an application image sample of the object to be classified in a preset application environment, and construct a function synthesis sample according to the application image sample.
It is easy to understand that, an application image sample of the object to be classified is obtained in a preset application environment, wherein the preset application environment can be the application environment of the object to be classified, and the preset classification number of the object to be classified is determined in the preset application environment; acquiring an application image sample of the object to be classified in the preset application environment according to the preset classification quantity; and constructing a function synthesis sample according to the distance information between the application image samples. The preset number of classifications is to take as many application image samples as the conditions allow.
Specifically, objects to be classified are shot in an application environment, each object in the objects to be classified needs to be shot and subjected to image classification marking, as many application image samples as possible are shot under conditions, function synthesis samples are constructed according to distance information between the application image samples, the function synthesis samples are used for training and verifying of a subsequent image classification model, and the more application image samples actually shot, the higher the accuracy of the image classification model obtained through training is.
It should be emphasized that, in order to further ensure the privacy and security of the background image samples and the foreground image samples, the background image samples and the foreground image samples may also be stored in nodes of a block chain.
And the fitting module 20 is configured to perform fitting according to the background image sample and the foreground image sample to obtain a synthesized training sample.
It should be noted that, fitting is performed according to the shot background image sample and the shot foreground image sample, so as to obtain a synthetic training sample. One feasible fitting method is to use a foreground image sample to cover a background image sample and perform fitting in a manner of performing background transparency processing near the edge to obtain a synthetic training sample, and the specific process may be as follows: segmenting the background image sample by a preset edge detection algorithm to obtain a target foreground image; classifying and combining the background image sample and the target foreground image according to a preset augmentation algorithm to obtain an initial synthesis training sample; and performing edge processing on the initial synthesis training sample according to the foreground image sample to obtain a synthesis training sample.
Specifically, a foreground environment is determined according to the preset background environment, the foreground environment can be a background environment as simple as possible, a foreground image sample of an object to be classified is collected in the background environment as simple as possible, the shot foreground image sample uses the background environment as simple as possible, a target foreground image can be obtained through a matting technology, for example, the background image sample can be segmented through a preset edge detection algorithm to obtain a target foreground image, namely, a foreground and a solid background of the background image sample are segmented through a simple edge detection algorithm to obtain the target foreground image.
Further, classifying and combining the background image sample and the target foreground image according to a preset augmentation algorithm to obtain an initial synthesis training sample; the preset augmentation algorithm may include such augmentation algorithms as blurring, rotation, distortion and color cast, and performs random combination after the background image sample and the target foreground image are classified, so as to obtain an initial synthesized training sample, make as many target foreground images as possible appear in as many background image sample combinations as possible, generally directly cover the target foreground images in partial areas of the background image sample, and perform classification labeling on the synthesized initial synthesized training sample according to different target foreground image classifications.
Further, edge processing is performed on the synthesized initial synthesis training sample, and generally, an object edge area of the synthesized initial synthesis training sample can be obtained according to the edge of the foreground image sample and the relative position on the background image sample, and transparency processing is performed on the object edge area to obtain the synthesis training sample.
And the building module 30 is configured to build a to-be-trained image classification model of the to-be-classified object based on a preset neural network.
It should be understood that the image classification model to be trained of the object to be classified is constructed based on a preset neural network, or the image classification model to be trained of the object to be classified is constructed according to an actual requirement model, the image classification model to be trained of the object to be classified may be constructed based on a classification algorithm of cognitive artificial design of a human engineer, the image classification model to be trained of the object to be classified may also be an image classification model trained based on a neural network, the synthesized training sample and the function synthesized sample in this embodiment may train the two types of image classification models to be trained, obtain a target image classification model, improve the precision of the image classification model, and the image classification model to be trained of the object to be classified may also be an image classification model to be trained obtained in other manners, which is not limited in this embodiment.
And the training module 40 is configured to train the image classification model to be trained through the synthetic training sample and the function synthetic sample to obtain a target image classification model.
It is easy to understand that, the image classification model to be trained is trained through the synthesis training sample and the function synthesis sample to obtain a target image classification model, and the training of the image classification model to be trained can adopt a variety of ways, and this embodiment is described by the following training ways: updating network parameters of an image classification model to be trained through the synthetic training sample to obtain a basic image classification model; obtaining classifier parameters for image classification, updating the classifier parameters of the basic image classification model according to the classifier parameters, and obtaining an updated basic image classification model; and updating the network parameters of the updated basic image classification model through the function synthesis sample to obtain a target image classification model.
It should be noted that, in this embodiment, other training manners may also be used to train the to-be-trained image classification model through a synthetic training sample and a function synthetic sample, so as to obtain the target image classification model, which is not limited in this embodiment. Image classification is an important part in the field of artificial intelligence, most application systems based on machine vision need an image classification model, attribute judgment is carried out on images collected by a camera, then other processing flows are carried out, and the classification capability of the image classification model is the basic capability in the field of machine vision. The embodiment enables the sample augmentation number to be increased exponentially, greatly improves the classification precision of the image classification model, and can be applied to the field of artificial intelligence such as machine vision.
In this embodiment, the obtaining module 10 is configured to obtain a background image sample and a foreground image sample of an object to be classified in a preset background environment, respectively; the obtaining module 10 is further configured to obtain an application image sample of the object to be classified in a preset application environment, and construct a function synthesis sample according to the application image sample; a fitting module 20, configured to perform fitting according to the background image sample and the foreground image sample to obtain a synthetic training sample; the building module 30 is used for building an image classification model to be trained of the object to be classified based on a preset neural network; and the training module 40 is configured to train the image classification model to be trained through the synthetic training sample and the function synthetic sample to obtain a target image classification model. This embodiment is compared in current training sample and is not enlarged or only enlarge based on simple affine transform, the image classification model precision that obtains is relatively poor, adopt background image sample and foreground image sample to handle the sample mode of enlarging of fitting again after handling respectively, can make sample number of enlarging obtain exponential increase, greatly promote image classification model's classification precision, it is poor to have solved the image classification model precision that prior art training sample is difficult to gather or too changeable leads to, the image classification model improves the technical problem that the efficiency of work is difficult to improve.
Based on the first embodiment of the training apparatus based on the image classification model, a second embodiment of the training apparatus based on the image classification model is provided.
In this embodiment, the obtaining module 10 is further configured to determine the number of image acquisitions according to a diversity condition of a preset background environment; acquiring background image samples of the object to be classified in the preset background environment according to the image acquisition quantity; and determining a foreground environment according to the preset background environment, and acquiring a foreground image sample in the foreground environment.
Further, the obtaining module 10 is further configured to determine a preset classification number of the objects to be classified in a preset application environment; acquiring an application image sample of the object to be classified in the preset application environment according to the preset classification quantity; and constructing a function synthesis sample according to the distance information between the application image samples.
Further, the fitting module 20 is further configured to segment the background image sample through a preset edge detection algorithm to obtain a target foreground image; classifying and combining the background image sample and the target foreground image according to a preset augmentation algorithm to obtain an initial synthesis training sample; and performing edge processing on the initial synthesis training sample according to the foreground image sample to obtain a synthesis training sample.
Further, the fitting module 20 is further configured to cover the foreground image sample on the background image sample, and obtain a relative position of the foreground image sample with respect to the background image sample; obtaining a synthetic edge area of an initial synthetic training sample according to the edge area of the foreground image sample and the relative position; performing transparency processing on the synthetic edge area to obtain an initial synthetic training sample after transparency processing; and taking the initial synthetic training sample after transparency processing as a synthetic training sample.
Further, the fitting module 20 is further configured to obtain a gray value of each pixel point of the background image sample in the synthesized edge region; acquiring gray values of all pixel points of the foreground image sample in the synthetic edge region; performing transparency processing on the gray value of each pixel point of the background image sample and the gray value of each pixel point of the foreground image sample based on preset transparency to obtain a pixel synthesis value of each pixel point in the synthesis edge area; and constructing an initial synthesis training sample after transparency processing according to the pixel synthesis value.
Further, the training module 40 is further configured to update network parameters of an image classification model to be trained through the synthetic training sample to obtain a basic image classification model; obtaining classifier parameters for image classification, updating the classifier parameters of the basic image classification model according to the classifier parameters, and obtaining an updated basic image classification model; and updating the network parameters of the updated basic image classification model through the function synthesis sample to obtain a target image classification model.
Other embodiments or specific implementation manners of the training device based on the image classification model of the present invention may refer to the above embodiments of the training method based on the image classification model, and are not described herein again.
The block chain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, a consensus mechanism, an encryption algorithm and the like. A block chain (Blockchain), which is essentially a decentralized database, is a series of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, so as to verify the validity (anti-counterfeiting) of the information and generate a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (e.g., a rom/ram, a magnetic disk, an optical disk) and includes instructions for enabling a terminal device (e.g., a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present invention.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (10)
1. An image classification model-based training method is characterized by comprising the following steps:
respectively acquiring a background image sample and a foreground image sample of an object to be classified in a preset background environment;
acquiring an application image sample of the object to be classified in a preset application environment, and constructing a function synthesis sample according to the application image sample;
fitting according to the background image sample and the foreground image sample to obtain a synthetic training sample;
constructing an image classification model to be trained of the object to be classified based on a preset neural network;
and training the image classification model to be trained through the synthetic training sample and the function synthetic sample to obtain a target image classification model.
2. The image classification model-based training method according to claim 1, wherein the step of respectively obtaining the background image sample and the foreground image sample of the object to be classified in a preset background environment comprises:
determining the image acquisition quantity according to the diversity condition of a preset background environment;
acquiring background image samples of the object to be classified in the preset background environment according to the image acquisition quantity;
and determining a foreground environment according to the preset background environment, and acquiring a foreground image sample in the foreground environment.
3. The image classification model-based training method according to claim 1, wherein the step of obtaining application image samples of the object to be classified in a preset application environment and constructing function synthesis samples according to the application image samples comprises:
determining a preset classification number of the objects to be classified in a preset application environment;
acquiring an application image sample of the object to be classified in the preset application environment according to the preset classification quantity;
and constructing a function synthesis sample according to the distance information between the application image samples.
4. The image classification model-based training method of claim 1, wherein the step of fitting the background image sample and the foreground image sample to obtain a synthetic training sample comprises:
segmenting the background image sample by a preset edge detection algorithm to obtain a target foreground image;
classifying and combining the background image sample and the target foreground image according to a preset augmentation algorithm to obtain an initial synthesis training sample;
and performing edge processing on the initial synthesis training sample according to the foreground image sample to obtain a synthesis training sample.
5. The image classification model-based training method of claim 4, wherein the step of performing edge processing on the initial synthetic training sample according to the foreground image sample to obtain a synthetic training sample comprises:
covering the foreground image sample on the background image sample, and acquiring the relative position of the foreground image sample relative to the background image sample;
obtaining a synthetic edge area of an initial synthetic training sample according to the edge area of the foreground image sample and the relative position;
performing transparency processing on the synthetic edge area to obtain an initial synthetic training sample after transparency processing;
and taking the initial synthetic training sample after transparency processing as a synthetic training sample.
6. The image classification model-based training method of claim 5, wherein the step of performing transparency processing on the synthesized edge region to obtain an initial synthesized training sample after transparency processing comprises:
acquiring gray values of all pixel points of the background image sample in the synthetic edge region;
acquiring gray values of all pixel points of the foreground image sample in the synthetic edge region;
performing transparency processing on the gray value of each pixel point of the background image sample and the gray value of each pixel point of the foreground image sample based on preset transparency to obtain a pixel synthesis value of each pixel point in the synthesis edge area;
and constructing an initial synthesis training sample after transparency processing according to the pixel synthesis value.
7. The image classification model-based training method according to any one of claims 1 to 6, wherein the step of training the image classification model to be trained by the synthetic training samples and the function synthetic samples to obtain a target image classification model comprises:
updating network parameters of an image classification model to be trained through the synthetic training sample to obtain a basic image classification model;
obtaining classifier parameters for image classification, updating the classifier parameters of the basic image classification model according to the classifier parameters, and obtaining an updated basic image classification model;
and updating the network parameters of the updated basic image classification model through the function synthesis sample to obtain a target image classification model.
8. An image classification model-based training device, characterized in that the image classification model-based training device comprises:
the acquisition module is used for respectively acquiring a background image sample and a foreground image sample of an object to be classified in a preset background environment;
the acquisition module is further used for acquiring an application image sample of the object to be classified in a preset application environment and constructing a function synthesis sample according to the application image sample;
the fitting module is used for fitting according to the background image sample and the foreground image sample to obtain a synthetic training sample;
the building module is used for building an image classification model to be trained of the object to be classified based on a preset neural network;
and the training module is used for training the image classification model to be trained through the synthetic training sample and the function synthetic sample to obtain a target image classification model.
9. An image classification model-based training apparatus, characterized in that the image classification model-based training apparatus comprises: a memory, a processor, and an image classification model-based training program stored on the memory and executable on the processor, the image classification model-based training program configured to implement the steps of the image classification model-based training method of any one of claims 1 to 7.
10. A storage medium, characterized in that the storage medium stores thereon an image classification model-based training program, which when executed by a processor implements the steps of the image classification model-based training method according to any one of claims 1 to 7.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010925560.3A CN112016630B (en) | 2020-09-03 | 2020-09-03 | Training method, device, equipment and storage medium based on image classification model |
| PCT/CN2020/125007 WO2021189847A1 (en) | 2020-09-03 | 2020-10-30 | Training method, apparatus and device based on image classification model, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010925560.3A CN112016630B (en) | 2020-09-03 | 2020-09-03 | Training method, device, equipment and storage medium based on image classification model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112016630A true CN112016630A (en) | 2020-12-01 |
| CN112016630B CN112016630B (en) | 2024-03-19 |
Family
ID=73515430
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010925560.3A Active CN112016630B (en) | 2020-09-03 | 2020-09-03 | Training method, device, equipment and storage medium based on image classification model |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN112016630B (en) |
| WO (1) | WO2021189847A1 (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113468906A (en) * | 2021-07-12 | 2021-10-01 | 深圳思谋信息科技有限公司 | Graphic code extraction model construction method, recognition method, device, equipment and medium |
| CN113762422A (en) * | 2021-11-09 | 2021-12-07 | 广州思德医疗科技有限公司 | Image training set composition method and system |
| WO2023020103A1 (en) * | 2021-08-17 | 2023-02-23 | 北京京东乾石科技有限公司 | Method and apparatus for updating target detection model |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114185429B (en) * | 2021-11-11 | 2024-03-26 | 杭州易现先进科技有限公司 | Gesture key point positioning or gesture estimating method, electronic device and storage medium |
| CN114359233B (en) * | 2022-01-07 | 2024-04-02 | 北京华云安信息技术有限公司 | Image segmentation model training method and device, electronic equipment and readable storage medium |
| CN114333038B (en) * | 2022-03-03 | 2022-06-07 | 百度在线网络技术（北京）有限公司 | Training method of object recognition model, object recognition method, device and equipment |
| CN114972404A (en) * | 2022-04-07 | 2022-08-30 | 咪咕文化科技有限公司 | High-precision image matting method and device, terminal equipment and computer readable storage medium |
| CN117152171A (en) * | 2022-05-24 | 2023-12-01 | 腾讯科技（深圳）有限公司 | Image processing method, apparatus, device, storage medium, and program product |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104715451A (en) * | 2015-03-11 | 2015-06-17 | 西安交通大学 | Seamless image fusion method based on consistent optimization of color and transparency |
| WO2018121690A1 (en) * | 2016-12-29 | 2018-07-05 | 北京市商汤科技开发有限公司 | Object attribute detection method and device, neural network training method and device, and regional detection method and device |
| CN108932735A (en) * | 2018-07-10 | 2018-12-04 | 广州众聚智能科技有限公司 | A method of generating deep learning sample |
| CN110516514A (en) * | 2018-05-22 | 2019-11-29 | 杭州海康威视数字技术股份有限公司 | A kind of modeling method and device of target detection model |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8903167B2 (en) * | 2011-05-12 | 2014-12-02 | Microsoft Corporation | Synthesizing training samples for object recognition |
| CN110197190B (en) * | 2018-02-27 | 2022-11-01 | 北京猎户星空科技有限公司 | Model training and object positioning method and device |
| CN108830780B (en) * | 2018-05-09 | 2020-09-01 | 京东数字科技控股有限公司 | Image processing method and device, electronic device and storage medium |
| CN111476284B (en) * | 2020-04-01 | 2023-12-26 | 杭州网易智企科技有限公司 | Image recognition model training and image recognition method and device and electronic equipment |
-
2020
- 2020-09-03 CN CN202010925560.3A patent/CN112016630B/en active Active
- 2020-10-30 WO PCT/CN2020/125007 patent/WO2021189847A1/en active Application Filing
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104715451A (en) * | 2015-03-11 | 2015-06-17 | 西安交通大学 | Seamless image fusion method based on consistent optimization of color and transparency |
| WO2018121690A1 (en) * | 2016-12-29 | 2018-07-05 | 北京市商汤科技开发有限公司 | Object attribute detection method and device, neural network training method and device, and regional detection method and device |
| CN110516514A (en) * | 2018-05-22 | 2019-11-29 | 杭州海康威视数字技术股份有限公司 | A kind of modeling method and device of target detection model |
| CN108932735A (en) * | 2018-07-10 | 2018-12-04 | 广州众聚智能科技有限公司 | A method of generating deep learning sample |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113468906A (en) * | 2021-07-12 | 2021-10-01 | 深圳思谋信息科技有限公司 | Graphic code extraction model construction method, recognition method, device, equipment and medium |
| CN113468906B (en) * | 2021-07-12 | 2024-03-26 | 深圳思谋信息科技有限公司 | Graphic code extraction model construction method, identification device, equipment and medium |
| WO2023020103A1 (en) * | 2021-08-17 | 2023-02-23 | 北京京东乾石科技有限公司 | Method and apparatus for updating target detection model |
| CN113762422A (en) * | 2021-11-09 | 2021-12-07 | 广州思德医疗科技有限公司 | Image training set composition method and system |
| CN113762422B (en) * | 2021-11-09 | 2022-04-05 | 广州思德医疗科技有限公司 | Image training set composition method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2021189847A1 (en) | 2021-09-30 |
| CN112016630B (en) | 2024-03-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112016630B (en) | Training method, device, equipment and storage medium based on image classification model | |
| JP5764238B2 (en) | Steel pipe internal corrosion analysis apparatus and steel pipe internal corrosion analysis method | |
| CN109829930B (en) | Face image processing method and device, computer equipment and readable storage medium | |
| RU2680765C1 (en) | Automated determination and cutting of non-singular contour of a picture on an image | |
| CN107204034B (en) | A kind of image processing method and terminal | |
| CN109978890A (en) | Target extraction method, device and terminal device based on image procossing | |
| CN101536078A (en) | Improving image masks | |
| US20160217263A1 (en) | Image processing apparatus, image processing method, image display system, and storage medium | |
| CN110930296A (en) | Image processing method, device, equipment and storage medium | |
| JP2017117019A (en) | Image processing device, image processing method, and program | |
| CN110599554A (en) | Method and device for identifying face skin color, storage medium and electronic device | |
| CN106097252A (en) | High spectrum image superpixel segmentation method based on figure Graph model | |
| CN111062863B (en) | Method, device, equipment and storage medium for binding 3D model with longitude and latitude coordinates | |
| CN114882306B (en) | Topography scale identification method and device, storage medium and electronic equipment | |
| CN114708237A (en) | Detection algorithm for hair health condition | |
| CN114677319A (en) | Stem cell distribution determination method and device, electronic equipment and storage medium | |
| CN104766068A (en) | Random walk tongue image extraction method based on multi-rule fusion | |
| CN109658360B (en) | Image processing method and device, electronic equipment and computer storage medium | |
| CN114693697A (en) | Image processing method, device, equipment and computer readable storage medium | |
| CN116597317A (en) | Remote sensing image change detection data generation method, device, equipment and medium | |
| JP5203159B2 (en) | Image processing method, image processing system, and image processing program | |
| CN116168106A (en) | Image processing method, device, computer equipment and storage medium | |
| Zhang et al. | A combined approach to single-camera-based lane detection in driverless navigation | |
| CN112571409B (en) | Robot control method based on visual SLAM, robot and medium | |
| CN113112475A (en) | Traditional Chinese medicine ear five-organ region segmentation method and device based on machine learning |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |