CN109146830A - For generating the method, apparatus, system and storage medium of training data - Google Patents
For generating the method, apparatus, system and storage medium of training data Download PDFInfo
- Publication number
- CN109146830A CN109146830A CN201810785130.9A CN201810785130A CN109146830A CN 109146830 A CN109146830 A CN 109146830A CN 201810785130 A CN201810785130 A CN 201810785130A CN 109146830 A CN109146830 A CN 109146830A
- Authority
- CN
- China
- Prior art keywords
- image
- training data
- foreground
- identified
- foreground image
- 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
- 238000012549 training Methods 0.000 title claims abstract description 90
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000003860 storage Methods 0.000 title claims abstract description 38
- 238000002156 mixing Methods 0.000 claims abstract description 49
- 238000004590 computer program Methods 0.000 claims description 25
- 239000011159 matrix material Substances 0.000 claims description 19
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 230000004927 fusion Effects 0.000 claims description 7
- 230000011218 segmentation Effects 0.000 claims description 7
- 238000001514 detection method Methods 0.000 abstract description 8
- 230000006870 function Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 241001465754 Metazoa Species 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 241000209140 Triticum Species 0.000 description 1
- 235000021307 Triticum Nutrition 0.000 description 1
- 230000000712 assembly Effects 0.000 description 1
- 238000000429 assembly Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010367 cloning Methods 0.000 description 1
- 230000009193 crawling Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000002844 melting Methods 0.000 description 1
- 230000008018 melting Effects 0.000 description 1
- 238000000465 moulding Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of for generating the method, apparatus, system and storage medium of training data, this method comprises: obtaining the image including scene as background image, and obtains the image including object to be identified as foreground image；And by predetermined position region that the foreground image is fitted in the background image to generate blending image, the blending image is as the training data.Method, apparatus, system and storage medium according to an embodiment of the present invention for generating training data using the image including object to be identified as foreground image by being fitted in the background image of all kinds of scenes, image of the object to be identified under all kinds of scenes can be quickly generated, so as to realize low cost, efficiently generate the training data for training to detection and/or the identification of object to be identified.
Description
Technical field
The present invention relates to object identification depth learning technology fields, relate more specifically to a kind of for generating training data
Method, apparatus, system and storage medium.
Background technique
The old process of the detection of object identification deep learning and recognizer is to need a large amount of acquisition objects in various scenes
Under picture, ensure that model to the generalization ability of object identification in this way (modelling effect is good).Current process is all logical
The picture that manually acquisition and network crawl to obtain object under all kinds of scenes is crossed, it is usually time-consuming and laborious, and also the project cycle is long.
Summary of the invention
The invention proposes a kind of schemes about for generating training data, by the figure that will include object to be identified
As being fitted in the background image of all kinds of scenes as foreground image, object to be identified can be quickly generated under all kinds of scenes
Image, so as to realize low cost, efficiently generate the training for training to detection and/or the identification of object to be identified
Data.The scheme proposed by the present invention about for generating training data is briefly described below, more details will be in subsequent combination
Attached drawing is described in a specific embodiment.
According to an aspect of the present invention, a kind of method for generating training data is provided, which comprises obtain packet
The image of scene is included as background image, and obtains the image including object to be identified as foreground image；And will be described before
The predetermined position region that scape image is fitted in the background image is to generate blending image, and the blending image is as the instruction
Practice data.
In one embodiment, the method also includes: before implementing the fitting, by the attribute of the foreground image
It is aligned with the attribute of the background image.
In one embodiment, the attribute includes color and/or angle.
In one embodiment, it is described by the foreground image be fitted in the predetermined position region in the background image with
Generating blending image is based on Poisson clone algorithm.
In one embodiment, it is described by the foreground image be fitted in the predetermined position region in the background image with
Generating blending image includes: the gradient fields for calculating separately the foreground image and the gradient fields of the background image；Based on described
The gradient fields of the gradient field computation of the gradient fields of foreground image and background image image to be generated；Based on the figure to be generated
The divergence of image to be generated described in the gradient field computation of picture；And the coefficient matrix of Poisson Reconstructed equation is solved, and based on described
The divergence of coefficient matrix and the image to be generated calculates the pixel value of each pixel of the image to be generated, to generate
Image to be generated is stated as the blending image.
In one embodiment, the image of the acquisition including object to be identified includes: that acquisition includes as foreground image
The image of the object to be identified；It and will include the object to be identified from the image including the object to be identified
Partial segmentation come out using as the foreground image.
In one embodiment, it is described obtain the foreground image further include obtain the foreground image identification information and
Attribute information, the training data further include that the identification information, attribute information and the foreground image of the foreground image exist
Location information in the blending image.
According to a further aspect of the invention, provide a kind of for generating the device of training data, described device includes: to obtain
Module for obtaining the image including scene as background image, and obtains the image including object to be identified as foreground picture
Picture；And fitting module, the predetermined position region for the foreground image to be fitted in the background image is melted with generating
Image is closed, the blending image is as the training data.
In one embodiment, the fitting module is also used to: before implementing the fitting, by the foreground image
The alignment of the attribute of attribute and the background image.
In one embodiment, the attribute includes color and/or angle.
In one embodiment, the foreground image is fitted in the pre-determined bit in the background image by the fitting module
It is based on Poisson clone algorithm that region, which is set, to generate blending image.
In one embodiment, the foreground image is fitted in the pre-determined bit in the background image by the fitting module
It includes: the gradient fields for calculating separately the foreground image and the gradient fields of the background image that region, which is set, to generate blending image；
The gradient fields of the gradient field computation image to be generated of gradient fields and the background image based on the foreground image；Based on described
The divergence of image to be generated described in the gradient field computation of image to be generated；And the coefficient matrix of Poisson Reconstructed equation is solved, and
Divergence based on the coefficient matrix and the image to be generated calculates the pixel value of each pixel of the image to be generated,
To generate the image to be generated as the blending image.
In one embodiment, the image of the module acquisition including object to be identified that obtain includes: as foreground image
Obtain the image including the object to be identified；And from the image including the object to be identified will include it is described to
Identify that the partial segmentation of object comes out using as the foreground image.
In one embodiment, it further includes the mark for obtaining the foreground image that the acquisition module, which obtains the foreground image,
Know information and attribute information, the training data further include the identification information of the foreground image, attribute information and it is described before
Location information of the scape image in the blending image.
Another aspect according to the present invention provides a kind of system for generating training data, and the system comprises storages
Device and processor are stored with the computer program run by the processor, the computer program on the storage device
The method described in any of the above embodiments for generating training data is executed when being run by the processor.
According to a further aspect of the present invention, a kind of storage medium is provided, is stored with computer program on the storage medium,
The computer program executes the method described in any of the above embodiments for generating training data at runtime.
Method, apparatus, system and storage medium according to an embodiment of the present invention for generating training data will be by that will wrap
The image for including object to be identified is fitted in the background image of all kinds of scenes as foreground image, and it is to be identified right to quickly generate
As the image under all kinds of scenes, so as to realize low cost, efficiently generate the inspection for training to object to be identified
The training data surveyed and/or identified.
Detailed description of the invention
The embodiment of the present invention is described in more detail in conjunction with the accompanying drawings, the above and other purposes of the present invention,
Feature and advantage will be apparent.Attached drawing is used to provide to further understand the embodiment of the present invention, and constitutes explanation
A part of book, is used to explain the present invention together with the embodiment of the present invention, is not construed as limiting the invention.In the accompanying drawings,
Identical reference label typically represents same parts or step.
Fig. 1 shows for realizing the method, apparatus according to an embodiment of the present invention for being used to generate training data, system and deposits
The schematic block diagram of the exemplary electronic device of storage media；
Fig. 2 shows according to an embodiment of the present invention for generating the schematic flow chart of the method for training data；
Fig. 3 shows according to an embodiment of the present invention for generating the schematic block diagram of the device of training data；And
Fig. 4 shows according to an embodiment of the present invention for generating the schematic block diagram of the system of training data.
Specific embodiment
In order to enable the object, technical solutions and advantages of the present invention become apparent, root is described in detail below with reference to accompanying drawings
According to example embodiments of the present invention.Obviously, described embodiment is only a part of the embodiments of the present invention, rather than this hair
Bright whole embodiments, it should be appreciated that the present invention is not limited by example embodiment described herein.Based on described in the present invention
The embodiment of the present invention, those skilled in the art's obtained all other embodiment in the case where not making the creative labor
It should all fall under the scope of the present invention.
Firstly, referring to Fig.1 come describe for realizing the embodiment of the present invention for generate training data method, apparatus,
System and the exemplary electronic device of storage medium 100.
As shown in Figure 1, electronic equipment 100 include one or more processors 102, it is one or more storage device 104, defeated
Enter device 106 and output device 108, these components (are not shown by the bindiny mechanism of bus system 110 and/or other forms
It interconnects out).It should be noted that the component and structure of electronic equipment 100 shown in FIG. 1 are illustrative, and not restrictive, root
According to needs, the electronic equipment also can have other assemblies and structure.
The processor 102 can be central processing unit (CPU), graphics processing unit (GPU) or have at data
The processing unit of reason ability and/or the other forms of instruction execution capability, and can control its in the electronic equipment 100
Its component is to execute desired function.
The storage device 104 may include one or more computer program products, and the computer program product can
To include various forms of computer readable storage mediums, such as volatile memory and/or nonvolatile memory.It is described easy
The property lost memory for example may include random access memory (RAM) and/or cache memory (cache) etc..It is described non-
Volatile memory for example may include read-only memory (ROM), hard disk, flash memory etc..In the computer readable storage medium
On can store one or more computer program instructions, processor 102 can run described program instruction, to realize hereafter institute
The client functionality (realized by processor) in the embodiment of the present invention stated and/or other desired functions.In the meter
Can also store various application programs and various data in calculation machine readable storage medium storing program for executing, for example, the application program use and/or
The various data etc. generated.
The input unit 106 can be the device that user is used to input instruction, and may include keyboard, mouse, wheat
One or more of gram wind and touch screen etc..
The output device 108 can export various information (such as image or sound) to external (such as user), and
It may include one or more of display, loudspeaker etc..
Illustratively, for realizing according to an embodiment of the present invention for generating the example of the method and apparatus of training data
Electronic equipment can be such as smart phone, tablet computer etc. mobile terminal.Illustratively, for realizing real according to the present invention
Applying the exemplary electronic device of the method and apparatus by generating training data of example may be any with based on computing capability
Calculate equipment.
In the following, reference Fig. 2 is described the method 200 according to an embodiment of the present invention for being used to generate training data.Such as Fig. 2 institute
Show, the method 200 for generating training data may include steps of:
In step S210, the image including scene is obtained as background image, and obtains the image including object to be identified
As foreground image.
In one example, template largely including all kinds of scenes can be prepared in advance, then pass through image collector
Set all kinds of scene backgrounds of (such as mobile phone, camera etc.) real scene shooting, with obtain include all kinds of scenes a large amount of background image.Another
In a example, can by interconnect network method (such as being crawled by network) obtain include all kinds of scenes a large amount of Background
Picture.In other examples, the image including scene can be obtained as background image by any other suitable means.Show
Example property, different background images may include the image under different scenes.Illustratively, different background images can also wrap
Include the image of different illumination, different angle etc. under Same Scene.
In one example, available includes object to be identified (such as pedestrian, face, animal, vehicle, text etc.)
Then partial segmentation including object to be identified is come out from the image including object to be identified and is used as foreground image by image.
For example, the image of object to be identified can be shot, then the extracting section of object to be identified in the captured image is come out, is picked
Except remaining background parts using obtain include the object to be identified image section as foreground image.In another example, institute
The foreground image of acquisition can be for only including the image of object to be identified.In other examples, it can also directly be obtained from any source
Take including or only including object to be identified image as foreground image.
In step S220, the predetermined position region that the foreground image is fitted in the background image is merged with generating
Image, the blending image is as the training data.
In embodiments of the present invention, since the position of the object to be identified in blending image is known, so not needing
The position of object to be identified is marked in blending image, which is used directly for training, so as to realize it is low at
Originally, the training data for training to detection and/or the identification of object to be identified is efficiently generated.
In one embodiment, (Seamless cloning) algorithm can be cloned based on Poisson come by the foreground image
The predetermined position region being fitted in the background image is to generate blending image.In one example, described by the prospect
The predetermined position region that image is fitted in the background image may include: to calculate separately the prospect to generate blending image
The gradient fields of the gradient fields of image and the background image；The ladder of gradient fields and the background image based on the foreground image
The gradient fields for spending field computation image to be generated (are replaced using the gradient fields of foreground image described predetermined in the background image
The gradient fields of the band of position)；The divergence of image to be generated described in gradient field computation based on the image to be generated；And it solves
The coefficient matrix of Poisson Reconstructed equation, and it is described to be generated based on the calculating of the divergence of the coefficient matrix and the image to be generated
The pixel value of each pixel of image, to generate the image to be generated as the blending image.Such as Poisson reconstruction side
Journey is Ax=b, and wherein A indicates that coefficient matrix, x indicate that the pixel value of each pixel in image to be generated, b indicate figure to be generated
The divergence of picture can solve x according to A and b.
In another example, it is described by the foreground image be fitted in the predetermined position region in the background image with
Generate blending image may include: calculate separately the foreground image gradient fields and the background image except the pre-determined bit
Set the gradient fields other than region；Gradient fields and the background image based on the foreground image remove the predetermined position region
The gradient fields of gradient field computation image to be generated in addition are (that is, add the background image for the gradient fields of the foreground image
The gradient fields in addition to the predetermined position region obtain the gradient fields of image to be generated)；Based on the image to be generated
Gradient field computation described in image to be generated divergence；And the coefficient matrix of Poisson Reconstructed equation is solved, and be based on the system
The divergence of matrix number and the image to be generated calculates the pixel value of each pixel of the image to be generated, described in generating
Image to be generated is as the blending image.
In other embodiments, the foreground image can also be fitted in using any other suitable method described
Predetermined position region in background image is to generate blending image.
It in one embodiment, can before the predetermined position region being fitted in foreground image in the background image
To be first aligned the attribute of the attribute of foreground image and background image, such as the color of foreground image, angle etc. can be belonged to
Property is each self-aligned with the attributes such as the color of background image, angle respectively, can make melting for background image and foreground image in this way
It is more natural to close effect.In other embodiments, other any suitable places can also be implemented to foreground image and background image
It manages to improve the syncretizing effect of the two.
It in an embodiment of the present invention, can be in the Background for the every background image obtained in step S210
The band of position for being bonded foreground image is preset as in, that is to say, that in the blending image formed after fitting, prospect
The band of position where image (i.e. object to be identified) is known.Further, in an embodiment of the present invention, before acquisition
When scape image, it can while obtaining the identification information even attribute information of foreground image.Wherein, the identification information of foreground image
It can indicate that the classification of object to be identified, such as instruction object to be identified are pedestrians, are faces, are animals, are vehicles, are texts
Or other classifications etc..Correspondingly, the attribute information of foreground image can indicate each attribute of object to be identified.For example,
When object to be identified is pedestrian, attribute information may include gender, age, clothing, posture, hair style of pedestrian etc.；When to
When identifying that object is face, whether attribute information may include expression, gender, the age, hair style, makes up；When to be identified right
When as vehicle, attribute information may include brand, vehicle, color, height, license plate number of vehicle etc..Therefore, foreground image
The identification information and attribute information of the location information and foreground image in blending image formed after fitting can be
Know, in this way, can be directly as the instruction for training to detection and/or the identification of object to be identified after forming blending image
Practice data to come using can further improve efficiency without being labeled.
Based on above description, the method according to an embodiment of the present invention for generating training data will be by that will include wait know
The image of other object is fitted in the background image of all kinds of scenes as foreground image, can quickly generate object to be identified each
Image under class scene, so as to realize low cost, efficiently generate for training to the detection of object to be identified and/or
The training data of identification.
The method according to an embodiment of the present invention for generating training data is described above exemplarily.Illustratively,
It is according to an embodiment of the present invention for generate training data method can with memory and processor unit or
It is realized in person's system.
In addition, the method according to an embodiment of the present invention for generating training data is deployed to intelligent hand in which can be convenient
In the mobile devices such as machine, tablet computer, personal computer.Alternatively, according to an embodiment of the present invention for generating training data
Method can also be deployed in server end (or cloud).Alternatively, according to an embodiment of the present invention for generating training data
Method can also be deployed in being distributed at server end (or cloud) and personal terminal.
The device for being used to generate training data of another aspect of the present invention offer is described below with reference to Fig. 3.Fig. 3 shows root
According to the schematic block diagram of the device 300 for generating training data of the embodiment of the present invention.
As shown in figure 3, the device 300 according to an embodiment of the present invention for generating training data includes obtaining module 310
With fitting module 320.The modules can execute the method for generating training data above in conjunction with Fig. 2 description respectively
Each step/function.Only the major function of each module of the device 300 for generating training data is described below,
And omit the detail content having been described above.
It obtains module 310 and is used to obtain the image including scene as background image, and obtaining includes object to be identified
Image is as foreground image.The predetermined position that fitting module 320 is used to be fitted in the foreground image in the background image
Region is to generate blending image, and the blending image is as the training data.Obtain module 310 and fitting module 320
It is realized with the program instruction that is stored in 102 Running storage device 104 of processor in electronic equipment as shown in Figure 1.
In one example, template largely including all kinds of scenes can be prepared in advance, then pass through image collector
Set all kinds of scene backgrounds of (such as mobile phone, camera etc.) real scene shooting, with obtained by acquisition module 310 include all kinds of scenes a large amount of back
Scape image.In another example, obtaining module 310 can be included by interconnecting network method (such as crawling by network)
The a large amount of background image of all kinds of scenes.In other examples, obtain module 310 can by any other suitable means come
The image including scene is obtained as background image.Illustratively, different background images may include the figure under different scenes
Picture.Illustratively, different background images can also include the image of different illumination, different angle etc. under Same Scene.
In one example, it includes object to be identified (such as pedestrian, face, animal, vehicle that it is available, which to obtain module 310,
, text etc.) image, then from include object to be identified image in will include that the partial segmentation of object to be identified comes out and makees
For foreground image.For example, can be using the image of image collecting device shooting object to be identified, then obtaining module 310 should
The extracting section of object to be identified comes out in captured image, rejects remaining background parts to obtain including the object to be identified
Image section is as foreground image.In another example, obtain foreground image acquired in module 310 can for only include to
Identify the image of object.In other examples, acquisition module 310 can also directly from any source, acquisition includes or only includes wait know
The image of other object is as foreground image.
In one embodiment, the foreground image can be fitted in by fitting module 320 based on Poisson clone algorithm
Predetermined position region in the background image is to generate blending image.In one example, the fitting module 320 will be described
The predetermined position region that foreground image is fitted in the background image with generate blending image may include: calculate separately it is described
The gradient fields of the gradient fields of foreground image and the background image；Gradient fields and the background image based on the foreground image
Gradient field computation image to be generated gradient fields (i.e. using foreground image gradient fields replace it is described in the background image
The gradient fields in predetermined position region)；The divergence of image to be generated described in gradient field computation based on the image to be generated；And
Solve the coefficient matrix of Poisson Reconstructed equation, and based on the divergence of the coefficient matrix and the image to be generated calculate it is described to
The pixel value for generating each pixel of image, to generate the image to be generated as the blending image.
In another example, the foreground image is fitted in pre- in the background image by the fitting module 320
Location area may include: the gradient fields for calculating separately the foreground image and the background image to generate blending image
Gradient fields in addition to the predetermined position region；Removing for gradient fields and the background image based on the foreground image is described
The gradient fields of gradient field computation image to be generated other than the region of predetermined position are (that is, the gradient fields of the foreground image are added
The gradient fields in addition to the predetermined position region of the background image obtain the gradient fields of image to be generated)；Based on institute
State the divergence of image to be generated described in the gradient field computation of image to be generated；And the coefficient matrix of Poisson Reconstructed equation is solved,
And the divergence based on the coefficient matrix and the image to be generated calculates the pixel of each pixel of the image to be generated
Value, to generate the image to be generated as the blending image.
In other embodiments, fitting module 320 can also use any other suitable method by the foreground picture
As the predetermined position region being fitted in the background image is to generate blending image.
In one embodiment, before the predetermined position region being fitted in foreground image in the background image, patch
Can first the attribute of the attribute of foreground image and background image be aligned by molding block 320, such as can be by the face of foreground image
The attributes such as color, angle are each self-aligned with the attributes such as the color of background image, angle respectively, can make background image with before in this way
The syncretizing effect of scape image is more natural.In other embodiments, fitting module 320 can also be to foreground image and background image
Implement other any suitable processing to improve the syncretizing effect of the two.
In an embodiment of the present invention, it for every acquired background image, can be set in advance in the background image
Determine the band of position for being bonded foreground image, that is to say, that in the blending image formed after fitting, foreground image (i.e. to
Identify object) where the band of position be known.Further, in an embodiment of the present invention, the acquisition of module 310 is being obtained
When foreground image, it can while obtaining the identification information even attribute information of foreground image.Wherein, the mark letter of foreground image
Breath can indicate that the classification of object to be identified, such as instruction object to be identified are pedestrians, are faces, are animals, are vehicles, are texts
Word or other classifications etc..Correspondingly, the attribute information of foreground image can indicate each attribute of object to be identified.Example
Such as, when object to be identified is pedestrian, attribute information may include gender, age, clothing, posture, hair style of pedestrian etc.；When
When object to be identified is face, whether attribute information may include expression, gender, the age, hair style, makes up；When to be identified
When object is vehicle, attribute information may include brand, vehicle, color, height, license plate number of vehicle etc..Therefore, foreground picture
As the identification information and attribute information of location information and foreground image in the blending image that is formed after fitting can be
It is known, in this way, after forming blending image can directly as training to the detection of object to be identified and/or identification
Training data comes using can further improve efficiency without being labeled.
Based on above description, the device according to an embodiment of the present invention for generating training data will be by that will include wait know
The image of other object is fitted in the background image of all kinds of scenes as foreground image, can quickly generate object to be identified each
Image under class scene, so as to realize low cost, efficiently generate for training to the detection of object to be identified and/or
The training data of identification.
Fig. 4 shows according to an embodiment of the present invention for generating the schematic block diagram of the system 400 of training data.For
The system 400 for generating training data includes storage device 410 and processor 420.
Wherein, the storage of storage device 410 is for realizing the method according to an embodiment of the present invention for generating training data
In corresponding steps program code.Program code of the processor 420 for being stored in Running storage device 410, to execute root
According to the corresponding steps of the method for generating training data of the embodiment of the present invention, and for realizing according to embodiments of the present invention
For generating the corresponding module in the device of training data.In addition, the system 400 for generating training data can also include
Image collecting device (not shown in FIG. 4), can be used for acquiring background image and foreground image.Certainly, image collector
It sets and is not required, can directly receive the input of background image and foreground image from other sources.
In one embodiment, make when said program code is run by processor 420 for generating training data
System 400 executes following steps: obtaining the image including scene as background image, and obtains the image including object to be identified
As foreground image；And by predetermined position region that the foreground image is fitted in the background image to generate fusion figure
Picture, the blending image is as the training data.
In one embodiment, also make when said program code is run by processor 420 for generating training data
System 400 execute following steps: before implementing the fitting, by the attribute of the foreground image and the background image
Attribute alignment.
In one embodiment, the attribute includes color and/or angle.
In one embodiment, make when said program code is run by processor 420 for generating training data
The predetermined position region that the foreground image is fitted in the background image that system 400 executes is to generate fusion figure
It seem based on Poisson clone algorithm.
In one embodiment, make when said program code is run by processor 420 for generating training data
The predetermined position region that the foreground image is fitted in the background image that system 400 executes is to generate fusion figure
Gradient fields as including: the gradient fields and background image for calculating separately the foreground image；Based on the foreground image
The gradient fields of the gradient field computation of gradient fields and background image image to be generated；Gradient fields based on the image to be generated
Calculate the divergence of the image to be generated；And solve the coefficient matrix of Poisson Reconstructed equation, and based on the coefficient matrix and
The divergence of the image to be generated calculates the pixel value of each pixel of the image to be generated, to generate the figure to be generated
As being used as the blending image.
In one embodiment, make when said program code is run by processor 420 for generating training data
The acquisition that system 400 executes includes that the image of object to be identified as foreground image includes: to obtain including described to be identified
The image of object；And by the partial segmentation including the object to be identified from the image including the object to be identified
Out using as the foreground image.
In one embodiment, make when said program code is run by processor 420 for generating training data
The acquisition foreground image that system 400 executes further includes the identification information and attribute information for obtaining the foreground image,
The training data further includes the identification information, attribute information and the foreground image of the foreground image in the fusion figure
Location information as in.
In addition, according to embodiments of the present invention, additionally providing a kind of storage medium, storing program on said storage
Instruction, when described program instruction is run by computer or processor for execute the embodiment of the present invention for generating trained number
According to method corresponding steps, and for realizing the phase according to an embodiment of the present invention for generating in the device of training data
Answer module.The storage medium for example may include the storage card of smart phone, the storage unit of tablet computer, personal computer
Hard disk, read-only memory (ROM), Erasable Programmable Read Only Memory EPROM (EPROM), portable compact disc read-only memory
(CD-ROM), any combination of USB storage or above-mentioned storage medium.The computer readable storage medium can be one
Any combination of a or multiple computer readable storage mediums.
In one embodiment, the computer program instructions may be implemented real according to the present invention when being run by computer
Each functional module of the device for generating training data of example is applied, and/or can be executed according to embodiments of the present invention
The method for generating training data.
In one embodiment, the computer program instructions make computer or place when being run by computer or processor
It manages device execution following steps: obtaining the image including scene as background image, and acquisition includes the image work of object to be identified
For foreground image；And by predetermined position region that the foreground image is fitted in the background image to generate fusion figure
Picture, the blending image is as the training data.
In one embodiment, the computer program instructions also make when being run by computer or processor computer or
Processor executes following steps: before implementing the fitting, by the category of the attribute of the foreground image and the background image
Property alignment.
In one embodiment, the attribute includes color and/or angle.
In one embodiment, the computer program instructions make computer or place when being run by computer or processor
The predetermined position region that the foreground image is fitted in the background image of device execution is managed to generate blending image
It is based on Poisson clone algorithm.
In one embodiment, the computer program instructions make computer or place when being run by computer or processor
The predetermined position region that the foreground image is fitted in the background image of device execution is managed to generate blending image
It include: the gradient fields for the gradient fields and background image for calculating separately the foreground image；Ladder based on the foreground image
Spend the gradient fields of the gradient field computation image to be generated of field and the background image；Based on the gradient fields of the image to be generated
Calculate the divergence of the image to be generated；And the coefficient matrix of Poisson Reconstructed equation is solved, and be based on the coefficient matrix and institute
State image to be generated divergence calculate the image to be generated each pixel pixel value, to generate the image to be generated
As the blending image.
In one embodiment, the computer program instructions make computer or place when being run by computer or processor
Reason device execute the acquisition include object to be identified image as foreground image include: obtain include the object to be identified
Image；And the partial segmentation including the object to be identified is come out from the image including the object to be identified
Using as the foreground image.
In one embodiment, the computer program instructions make computer or place when being run by computer or processor
Managing the acquisition foreground image that device executes further includes the identification information and attribute information for obtaining the foreground image, described
Training data further includes the identification information, attribute information and the foreground image of the foreground image in the blending image
Location information.
Each module in device according to an embodiment of the present invention for generating training data can be by according to the present invention
The processor of the electronic equipment for generating training data of embodiment runs the computer program instructions stored in memory
It realizes, or can be stored in the computer readable storage medium of computer program product according to an embodiment of the present invention
Realization when computer instruction is run by computer.
Method, apparatus, system and storage medium according to an embodiment of the present invention for generating training data will be by that will wrap
The image for including object to be identified is fitted in the background image of all kinds of scenes as foreground image, and it is to be identified right to quickly generate
As the image under all kinds of scenes, so as to realize low cost, efficiently generate the inspection for training to object to be identified
The training data surveyed and/or identified.
According to embodiments of the present invention, additionally provide a kind of computer program, the computer program can store beyond the clouds or
On local storage medium.When the computer program is run by computer or processor for executing the use of the embodiment of the present invention
In the corresponding steps for the method for generating training data, and for realizing according to an embodiment of the present invention for generating training data
Device in corresponding module.
Although describing example embodiment by reference to attached drawing here, it should be understood that above example embodiment are only exemplary
, and be not intended to limit the scope of the invention to this.Those of ordinary skill in the art can carry out various changes wherein
And modification, it is made without departing from the scope of the present invention and spiritual.All such changes and modifications are intended to be included in appended claims
Within required the scope of the present invention.
Those of ordinary skill in the art may be aware that list described in conjunction with the examples disclosed in the embodiments of the present disclosure
Member and algorithm steps can be realized with the combination of electronic hardware or computer software and electronic hardware.These functions are actually
It is implemented in hardware or software, the specific application and design constraint depending on technical solution.Professional technician
Each specific application can be used different methods to achieve the described function, but this realization is it is not considered that exceed
The scope of the present invention.
In several embodiments provided herein, it should be understood that disclosed device and method can pass through it
Its mode is realized.For example, apparatus embodiments described above are merely indicative, for example, the division of the unit, only
Only a kind of logical function partition, there may be another division manner in actual implementation, such as multiple units or components can be tied
Another equipment is closed or is desirably integrated into, or some features can be ignored or not executed.
In the instructions provided here, numerous specific details are set forth.It is to be appreciated, however, that implementation of the invention
Example can be practiced without these specific details.In some instances, well known method, structure is not been shown in detail
And technology, so as not to obscure the understanding of this specification.
Similarly, it should be understood that in order to simplify the present invention and help to understand one or more of the various inventive aspects,
To in the description of exemplary embodiment of the present invention, each feature of the invention be grouped together into sometimes single embodiment, figure,
Or in descriptions thereof.However, the method for the invention should not be construed to reflect an intention that i.e. claimed
The present invention claims features more more than feature expressly recited in each claim.More precisely, such as corresponding power
As sharp claim reflects, inventive point is that the spy of all features less than some disclosed single embodiment can be used
Sign is to solve corresponding technical problem.Therefore, it then follows thus claims of specific embodiment are expressly incorporated in this specific
Embodiment, wherein each, the claims themselves are regarded as separate embodiments of the invention.
It will be understood to those skilled in the art that any combination pair can be used other than mutually exclusive between feature
All features disclosed in this specification (including adjoint claim, abstract and attached drawing) and so disclosed any method
Or all process or units of equipment are combined.Unless expressly stated otherwise, this specification (is wanted including adjoint right
Ask, make a summary and attached drawing) disclosed in each feature can be replaced with an alternative feature that provides the same, equivalent, or similar purpose.
In addition, it will be appreciated by those of skill in the art that although some embodiments described herein include other embodiments
In included certain features rather than other feature, but the combination of the feature of different embodiments mean it is of the invention
Within the scope of and form different embodiments.For example, in detail in the claims, embodiment claimed it is one of any
Can in any combination mode come using.
Various component embodiments of the invention can be implemented in hardware, or to run on one or more processors
Software module realize, or be implemented in a combination thereof.It will be understood by those of skill in the art that can be used in practice
Microprocessor or digital signal processor (DSP) realize some or all of some modules according to an embodiment of the present invention
Function.The present invention is also implemented as some or all program of device (examples for executing method as described herein
Such as, computer program and computer program product).It is such to realize that program of the invention can store in computer-readable medium
On, or may be in the form of one or more signals.Such signal can be downloaded from an internet website to obtain, or
Person is provided on the carrier signal, or is provided in any other form.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and ability
Field technique personnel can be designed alternative embodiment without departing from the scope of the appended claims.In the claims,
Any reference symbol between parentheses should not be configured to limitations on claims.Word "comprising" does not exclude the presence of not
Element or step listed in the claims.Word "a" or "an" located in front of the element does not exclude the presence of multiple such
Element.The present invention can be by means of including the hardware of several different elements and being come by means of properly programmed computer real
It is existing.In the unit claims listing several devices, several in these devices can be through the same hardware branch
To embody.The use of word first, second, and third does not indicate any sequence.These words can be explained and be run after fame
Claim.
The above description is merely a specific embodiment or to the explanation of specific embodiment, protection of the invention
Range is not limited thereto, and anyone skilled in the art in the technical scope disclosed by the present invention, can be easily
Expect change or replacement, should be covered by the protection scope of the present invention.Protection scope of the present invention should be with claim
Subject to protection scope.
Claims (10)
1. a kind of method for generating training data, which is characterized in that the described method includes:
The image including scene is obtained as background image, and obtains the image including object to be identified as foreground image；With
And
By predetermined position region that the foreground image is fitted in the background image to generate blending image, the fusion figure
As being used as the training data.
2. the method according to claim 1, wherein the method also includes:
Before implementing the fitting, the attribute of the attribute of the foreground image and the background image is aligned.
3. according to the method described in claim 2, it is characterized in that, the attribute includes color and/or angle.
4. method according to claim 1 or 2, which is characterized in that described that the foreground image is fitted in the background
Predetermined position region in image is based on Poisson clone algorithm to generate blending image.
5. according to the method described in claim 4, it is characterized in that, described be fitted in the background image for the foreground image
In predetermined position region include: to generate blending image
Calculate separately the gradient fields of the foreground image and the gradient fields of the background image；
The gradient fields of the gradient field computation image to be generated of gradient fields and the background image based on the foreground image；
The divergence of image to be generated described in gradient field computation based on the image to be generated；And
The coefficient matrix of Poisson Reconstructed equation is solved, and institute is calculated based on the divergence of the coefficient matrix and the image to be generated
The pixel value for stating each pixel of image to be generated, to generate the image to be generated as the blending image.
6. method according to claim 1 or 2, which is characterized in that described to obtain the image conduct including object to be identified
Foreground image includes:
Obtain the image including the object to be identified；And
From the image including the object to be identified by include the object to be identified partial segmentation come out using as
The foreground image.
7. method according to claim 1 or 2, which is characterized in that the acquisition foreground image further includes obtaining institute
The identification information and attribute information of foreground image are stated, the training data further includes the identification information of the foreground image, attribute
The location information of information and the foreground image in the blending image.
8. a kind of for generating the device of training data, which is characterized in that described device includes:
Module is obtained, for obtaining the image including scene as background image, and the image including object to be identified is obtained and makees
For foreground image；And
It is bonded module, the predetermined position region for the foreground image to be fitted in the background image is to generate fusion figure
Picture, the blending image is as the training data.
9. a kind of system for generating training data, which is characterized in that the system comprises storage devices and processor, described
The computer program run by the processor is stored on storage device, the computer program is run by the processor
Method for generate training data of the Shi Zhihang as described in any one of claim 1-7.
10. a kind of storage medium, which is characterized in that be stored with computer program, the computer program on the storage medium
The method for generating training data as described in any one of claim 1-7 is executed at runtime.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810785130.9A CN109146830A (en) | 2018-07-17 | 2018-07-17 | For generating the method, apparatus, system and storage medium of training data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810785130.9A CN109146830A (en) | 2018-07-17 | 2018-07-17 | For generating the method, apparatus, system and storage medium of training data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109146830A true CN109146830A (en) | 2019-01-04 |
Family
ID=64800889
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810785130.9A Pending CN109146830A (en) | 2018-07-17 | 2018-07-17 | For generating the method, apparatus, system and storage medium of training data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109146830A (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109727264A (en) * | 2019-01-10 | 2019-05-07 | 南京旷云科技有限公司 | Image generating method, the training method of neural network, device and electronic equipment |
| CN110033463A (en) * | 2019-04-12 | 2019-07-19 | 腾讯科技（深圳）有限公司 | A kind of foreground data generates and its application method, relevant apparatus and system |
| CN110414480A (en) * | 2019-08-09 | 2019-11-05 | 威盛电子股份有限公司 | Training image production method and electronic device |
| CN110503146A (en) * | 2019-08-21 | 2019-11-26 | 杭州比智科技有限公司 | Data enhancement methods, calculate equipment and computer storage medium at device |
| CN111462162A (en) * | 2019-01-18 | 2020-07-28 | 上海大学 | Foreground segmentation algorithm for specific class of pictures |
| CN111709873A (en) * | 2020-05-27 | 2020-09-25 | 北京百度网讯科技有限公司 | Training method and device of image conversion model generator |
| CN111753836A (en) * | 2019-08-27 | 2020-10-09 | 北京京东尚科信息技术有限公司 | Character recognition method and device, computer readable medium and electronic equipment |
| CN111783874A (en) * | 2020-06-29 | 2020-10-16 | 联想(北京)有限公司 | Sample generation method, sample training method and device |
| CN111899204A (en) * | 2020-07-30 | 2020-11-06 | 平安科技（深圳）有限公司 | Vehicle damage detection data synthesis method and device and storage medium |
| CN112102315A (en) * | 2020-11-02 | 2020-12-18 | 腾讯科技（深圳）有限公司 | Medical image processing method, medical image processing device, computer equipment and storage medium |
| CN113688887A (en) * | 2021-08-13 | 2021-11-23 | 百度在线网络技术（北京）有限公司 | Training and image recognition method and device of image recognition model |
| CN113902956A (en) * | 2021-09-30 | 2022-01-07 | 北京百度网讯科技有限公司 | Training method of fusion model, image fusion method, device, equipment and medium |
| CN114612744A (en) * | 2022-03-10 | 2022-06-10 | 平安科技（深圳）有限公司 | Detection model training method, vehicle damage detection method and terminal equipment |
| CN114820609A (en) * | 2022-06-29 | 2022-07-29 | 南昌大学 | Photovoltaic module EL image defect detection method |
| CN117523341A (en) * | 2023-11-23 | 2024-02-06 | 中船(北京)智能装备科技有限公司 | Deep learning training image sample generation method, device and equipment |
| US11967081B2 (en) | 2019-01-22 | 2024-04-23 | Mitsubishi Electric Corporation | Information processing apparatus, non-transitory computer-readable storage medium, and information processing method |
| CN117523341B (en) * | 2023-11-23 | 2024-06-21 | 中船(北京)智能装备科技有限公司 | Deep learning training image sample generation method, device and equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107154046A (en) * | 2017-04-06 | 2017-09-12 | 南京邮电大学 | A kind of method of video background processing and secret protection |
| CN107767391A (en) * | 2017-11-02 | 2018-03-06 | 北京奇虎科技有限公司 | Landscape image processing method, device, computing device and computer-readable storage medium |
| CN107808373A (en) * | 2017-11-15 | 2018-03-16 | 北京奇虎科技有限公司 | Sample image synthetic method, device and computing device based on posture |
| CN108156370A (en) * | 2017-12-07 | 2018-06-12 | Tcl移动通信科技（宁波）有限公司 | By the use of local picture as the photographic method of background, storage medium and mobile terminal |
-
2018
- 2018-07-17 CN CN201810785130.9A patent/CN109146830A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107154046A (en) * | 2017-04-06 | 2017-09-12 | 南京邮电大学 | A kind of method of video background processing and secret protection |
| CN107767391A (en) * | 2017-11-02 | 2018-03-06 | 北京奇虎科技有限公司 | Landscape image processing method, device, computing device and computer-readable storage medium |
| CN107808373A (en) * | 2017-11-15 | 2018-03-16 | 北京奇虎科技有限公司 | Sample image synthetic method, device and computing device based on posture |
| CN108156370A (en) * | 2017-12-07 | 2018-06-12 | Tcl移动通信科技（宁波）有限公司 | By the use of local picture as the photographic method of background, storage medium and mobile terminal |
Non-Patent Citations (2)
| Title |
|---|
| 《美术大观》编辑部: "《中国美术教育学术论丛 艺术设计卷 6》", 31 October 2016 * |
| 崔丽 等: "《Matlab小波分析与应用：30个案例分析》", 30 June 2016 * |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109727264A (en) * | 2019-01-10 | 2019-05-07 | 南京旷云科技有限公司 | Image generating method, the training method of neural network, device and electronic equipment |
| CN111462162A (en) * | 2019-01-18 | 2020-07-28 | 上海大学 | Foreground segmentation algorithm for specific class of pictures |
| CN111462162B (en) * | 2019-01-18 | 2023-07-21 | 上海大学 | Foreground segmentation algorithm for specific class pictures |
| US11967081B2 (en) | 2019-01-22 | 2024-04-23 | Mitsubishi Electric Corporation | Information processing apparatus, non-transitory computer-readable storage medium, and information processing method |
| CN110033463B (en) * | 2019-04-12 | 2021-06-04 | 腾讯科技（深圳）有限公司 | Foreground data generation and application method thereof, and related device and system |
| CN110033463A (en) * | 2019-04-12 | 2019-07-19 | 腾讯科技（深圳）有限公司 | A kind of foreground data generates and its application method, relevant apparatus and system |
| US11961237B2 (en) | 2019-04-12 | 2024-04-16 | Tencent Technology (Shenzhen) Company Limited | Foreground data generation method and method for applying same, related apparatus, and system |
| CN110414480A (en) * | 2019-08-09 | 2019-11-05 | 威盛电子股份有限公司 | Training image production method and electronic device |
| CN110503146A (en) * | 2019-08-21 | 2019-11-26 | 杭州比智科技有限公司 | Data enhancement methods, calculate equipment and computer storage medium at device |
| CN111753836A (en) * | 2019-08-27 | 2020-10-09 | 北京京东尚科信息技术有限公司 | Character recognition method and device, computer readable medium and electronic equipment |
| CN111709873A (en) * | 2020-05-27 | 2020-09-25 | 北京百度网讯科技有限公司 | Training method and device of image conversion model generator |
| CN111783874A (en) * | 2020-06-29 | 2020-10-16 | 联想(北京)有限公司 | Sample generation method, sample training method and device |
| WO2021139357A1 (en) * | 2020-07-30 | 2021-07-15 | 平安科技（深圳）有限公司 | Vehicle damage check data synthesis method and system, and apparatus and storage medium |
| CN111899204A (en) * | 2020-07-30 | 2020-11-06 | 平安科技（深圳）有限公司 | Vehicle damage detection data synthesis method and device and storage medium |
| CN111899204B (en) * | 2020-07-30 | 2024-04-09 | 平安科技（深圳）有限公司 | Vehicle loss detection data synthesis method, device and storage medium |
| CN112102315A (en) * | 2020-11-02 | 2020-12-18 | 腾讯科技（深圳）有限公司 | Medical image processing method, medical image processing device, computer equipment and storage medium |
| CN112102315B (en) * | 2020-11-02 | 2021-02-19 | 腾讯科技（深圳）有限公司 | Medical image processing method, medical image processing device, computer equipment and storage medium |
| CN113688887A (en) * | 2021-08-13 | 2021-11-23 | 百度在线网络技术（北京）有限公司 | Training and image recognition method and device of image recognition model |
| CN113902956A (en) * | 2021-09-30 | 2022-01-07 | 北京百度网讯科技有限公司 | Training method of fusion model, image fusion method, device, equipment and medium |
| WO2023050868A1 (en) * | 2021-09-30 | 2023-04-06 | 北京百度网讯科技有限公司 | Method and apparatus for training fusion model, image fusion method and apparatus, and device and medium |
| CN114612744A (en) * | 2022-03-10 | 2022-06-10 | 平安科技（深圳）有限公司 | Detection model training method, vehicle damage detection method and terminal equipment |
| CN114820609A (en) * | 2022-06-29 | 2022-07-29 | 南昌大学 | Photovoltaic module EL image defect detection method |
| CN117523341A (en) * | 2023-11-23 | 2024-02-06 | 中船(北京)智能装备科技有限公司 | Deep learning training image sample generation method, device and equipment |
| CN117523341B (en) * | 2023-11-23 | 2024-06-21 | 中船(北京)智能装备科技有限公司 | Deep learning training image sample generation method, device and equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109146830A (en) | For generating the method, apparatus, system and storage medium of training data | |
| CN110084874B (en) | Image style migration for three-dimensional models | |
| US20220383649A1 (en) | System and method for facilitating graphic-recognition training of a recognition model | |
| JP2022513272A (en) | Training A method and system for automatically generating mass training datasets from 3D models of deep learning networks | |
| CN106203305A (en) | Human face in-vivo detection method and device | |
| CN110363213B (en) | Method and system for cognitive analysis and classification of garment images | |
| US9262853B2 (en) | Virtual scene generation based on imagery | |
| CN108876833A (en) | Image processing method, image processing apparatus and computer readable storage medium | |
| CN108875932A (en) | Image-recognizing method, device and system and storage medium | |
| CN108037823B (en) | Information recommendation method, Intelligent mirror and computer readable storage medium | |
| CN108876804A (en) | It scratches as model training and image are scratched as methods, devices and systems and storage medium | |
| CN109919211A (en) | Commodity recognition method, device, system and computer-readable medium | |
| CN108875523A (en) | Human synovial point detecting method, device, system and storage medium | |
| US20210027514A1 (en) | Method and system for creating animal type avatar using human face | |
| US11256958B1 (en) | Training with simulated images | |
| CN108961149A (en) | Image processing method, device and system and storage medium | |
| CN108875515A (en) | Face identification method, device, system, storage medium and capture machine | |
| CN106132341B (en) | The prediction of shape | |
| CN108875731A (en) | Target identification method, device, system and storage medium | |
| CN108875483A (en) | Image detecting method, device, system and computer-readable medium | |
| CN106651973A (en) | Image structuring method and device | |
| CN111008935A (en) | Face image enhancement method, device, system and storage medium | |
| CN107959798A (en) | Video data real-time processing method and device, computing device | |
| CN111311480A (en) | Image fusion method and device | |
| CN110490058A (en) | Training method, device, system and the computer-readable medium of pedestrian detection model |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190104 |