CN110349161A - Image partition method, device, electronic equipment and storage medium - Google Patents
Image partition method, device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN110349161A CN110349161A CN201910621621.4A CN201910621621A CN110349161A CN 110349161 A CN110349161 A CN 110349161A CN 201910621621 A CN201910621621 A CN 201910621621A CN 110349161 A CN110349161 A CN 110349161A
- Authority
- CN
- China
- Prior art keywords
- category
- image
- probability matrix
- split
- matrix
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
- G06T2207/30201—Face
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the present disclosure discloses a kind of image partition method, device, electronic equipment and storage medium, and method includes: to obtain image to be split；By the multi-class identification model of the image input training in advance to be split, first category probability matrix is obtained according to the output result information of the first object layer of the multi-class identification model, and second category probability matrix is obtained according to the output result information of the second destination layer of the multi-class identification model, wherein, the second category belongs to the subclass of the first category, corresponding position pixel belongs to the probability value of first category in image to be split described in the element representation of the first category probability matrix, corresponding position pixel belongs to the probability value of second category in image to be split described in the element representation of the second category probability matrix.The technical solution of the embodiment of the present disclosure quickly can carry out three Classification Semantics segmentations to image, can be improved image segmentation efficiency.
Description
Technical field
The embodiment of the present disclosure is related to machine learning techniques field, and in particular to a kind of image partition method, device, electronics are set
Standby and storage medium.
Background technique
Currently, (such as retouching desk figure, U.S. face are taken pictures) needs to carry out image in semantic segmentation, figure in various application scenarios
The purpose of picture segmentation is classified to each pixel in image, and as each pixel stamps class label.Carry out image point
When cutting, facing need the problem of being categorized into three classes sometimes, such as carrying out U.S. face and take pictures, spot acne on the face and general should be removed
Logical mole etc. influences beautiful element, and needs to retain more classical mole.
For such as above-mentioned this image segmentation problem, way relatively broad at present is usually to mark training sample
Model training is carried out for three kinds of labels, such as is labeled as 0,1,2, belong to these three types of probability according to each pixel and generates table respectively
Levy three probability graphs of each type.Its model training stage complexity of this method is undoubtedly relatively high, and the difficulty of calculating also compares
It is larger, therefore the efficiency for carrying out image segmentation is lower.
Summary of the invention
In view of this, the embodiment of the present disclosure provides a kind of image partition method, device, electronic equipment and storage medium, with
It realizes enough quickly to image three Classification Semantics segmentations of progress.
Other characteristics and advantages of the embodiment of the present disclosure will be apparent from by the following detailed description, or partially by
The practice of the embodiment of the present disclosure and acquistion.
In a first aspect, the embodiment of the present disclosure provides a kind of image partition method, comprising:
Obtain image to be split；
By the multi-class identification model of the image input training in advance to be split, according to the multi-class identification model
The output result information of first object layer obtains first category probability matrix and according to the second of the multi-class identification model
The output result information of destination layer obtains second category probability matrix, wherein the second category belongs to the first category
The size of subclass, the first category probability matrix, the second category probability matrix and the image to be split is homogeneous
Together, corresponding position pixel belongs to the general of first category in image to be split described in the element representation of the first category probability matrix
Rate value, corresponding position pixel belongs to second category in image to be split described in the element representation of the second category probability matrix
Probability value.
Second aspect, the embodiment of the present disclosure additionally provide a kind of image segmentation device, comprising:
Image acquisition unit to be split, for obtaining image to be split；
Classification recognition unit, the multi-class identification model for training the image input to be split in advance, according to institute
The output result information for stating the first object layer of multi-class identification model obtains first category probability matrix and according to described more
The output result information of second destination layer of classification identification model obtains second category probability matrix, wherein the second category
Belong to the subclass of the first category, the first category probability matrix, the second category probability matrix and it is described to
The size of segmented image is all the same, corresponding position picture in image to be split described in the element representation of the first category probability matrix
Element belongs to the probability value of first category, corresponding position in image to be split described in the element representation of the second category probability matrix
Pixel belongs to the probability value of second category.
The third aspect, the embodiment of the present disclosure additionally provide a kind of electronic equipment, comprising:
One or more processors；
Memory, for storing one or more programs；
When one or more of programs are executed by one or more of processors, so that one or more of processing
Device realizes the instruction such as any one of first aspect the method.
Fourth aspect, the embodiment of the present disclosure additionally provide a kind of computer readable storage medium, are stored thereon with computer
Program is realized when the computer program is executed by processor such as the step of any one of first aspect the method.
The embodiment of the present disclosure is by inputting the multi-class identification model trained in advance for image to be split, according to the multiclass
The output result information of the first object layer of other identification model obtains first category probability matrix and according to the multi-class knowledge
The output result information of second destination layer of other model obtains second category probability matrix, quickly can carry out three classification to image
Semantic segmentation can be improved image segmentation efficiency.
Detailed description of the invention
It, below will be to institute in embodiment of the present disclosure description in order to illustrate more clearly of the technical solution in the embodiment of the present disclosure
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only in the embodiment of the present disclosure
A part of the embodiment for those of ordinary skill in the art without creative efforts, can also basis
The content of the embodiment of the present disclosure and these attached drawings obtain other attached drawings.
Fig. 1 is a kind of flow diagram for image partition method that the embodiment of the present disclosure provides；
Fig. 2 is the flow diagram for another image partition method that the embodiment of the present disclosure provides；
Fig. 3 is a kind of flow diagram of the training method for multi-class identification model that the embodiment of the present disclosure provides；
Fig. 4 is a kind of structural schematic diagram for image segmentation device that the embodiment of the present disclosure provides；
Fig. 5 is the structural schematic diagram for another image segmentation device that the embodiment of the present disclosure provides；
Fig. 6 is a kind of structural schematic diagram of the training device for multi-class identification model that the embodiment of the present disclosure provides；
Fig. 7 shows the structural schematic diagram for being suitable for the electronic equipment for being used to realize the embodiment of the present disclosure.
Specific embodiment
The technical issues of to solve the embodiment of the present disclosure, the technical solution of use and the technical effect that reaches are more clear
Chu is described in further detail, it is clear that described implementation below in conjunction with technical solution of the attached drawing to the embodiment of the present disclosure
Example is only a part of the embodiment in the embodiment of the present disclosure, instead of all the embodiments.Based in the embodiment of the present disclosure
Embodiment, those skilled in the art's every other embodiment obtained without creative efforts, belongs to
The range of embodiment of the present disclosure protection.
It should be noted that term " system " and " network " are often used interchangeably herein in the embodiment of the present disclosure.
The "and/or" mentioned in the embodiment of the present disclosure refers to " include one or more related listed items any and all combinations.
The specification and claims of the disclosure and term " first " in attached drawing, " second " etc. be for distinguishing different objects, and
It is not intended to limit particular order.
Also it should be noted that, following each embodiments can be individually performed in the embodiment of the present disclosure, between each embodiment
Can also be combined with each other execution, and the embodiment of the present disclosure is not specifically limited this.
Further illustrate the technical solution of the embodiment of the present disclosure below with reference to the accompanying drawings and specific embodiments.
Fig. 1 shows a kind of flow diagram of image partition method of embodiment of the present disclosure offer, and the present embodiment can fit
The case where for carrying out the segmentation of three Classification Semantics to image, this method can be by the image segmentation devices that are configured in electronic equipment
It executes, as shown in Figure 1, image partition method described in the present embodiment includes:
In step s 110, image to be split is obtained.
For example, the image to be split is facial image, the first category is spot acne mole classification, and the second category is
Mole classification with setting feature, such as belongs to classical mole, this image partition method can be used for dispelling spot acne on the face and common
Mole and retain classical mole.
By taking the image to be split is facial image as an example, the operation for obtaining image to be split includes: to obtain including face
Original image；Facial image to be processed in the original image is obtained as the image to be split.
Wherein, facial image to be processed in original image is obtained, a variety of methods can be used, for example, can will be described original
Image carries out face contour and analyzes to obtain face mask information, is cut out according to the face mask information to the original image
It cuts to obtain facial image to be processed.
Wherein, the original image including face, can be the image shot in advance, be also possible to obtain camera shooting in real time
Collected photo, by the photo be cached in buffer area as it is described include face original image.It, can for the former
It is used to carry out later period reparation to picture using the technical solution of the present embodiment that the technical solution of this implementation can be used for the latter
Real-time perfoming filter shooting, to shoot the photo or picture recording having dispelled spot acne and common mole and remained classical mole.
In the step s 120, by the multi-class identification model of the image input training in advance to be split, according to described more
The output result information of the first object layer of classification identification model obtains first category probability matrix and according to described multi-class
The output result information of second destination layer of identification model obtains second category probability matrix.
Wherein, the second category belongs to the subclass of the first category, the first category probability matrix, described
Two category probability matrix and the size of the image to be split are all the same, the element representation of the first category probability matrix
Corresponding position pixel belongs to the probability value of first category, the list of elements of the second category probability matrix in the image to be split
Show that corresponding position pixel in the image to be split belongs to the probability value of second category.
The embodiment of the present disclosure is by inputting the multi-class identification model trained in advance for image to be split, according to the multiclass
The output result information of the first object layer of other identification model obtains first category probability matrix and according to the multi-class knowledge
The output result information of second destination layer of other model obtains second category probability matrix, quickly can carry out three classification to image
Segmentation, can be improved image segmentation efficiency.
Fig. 2 shows the embodiment of the present disclosure provide another image partition method flow diagram, the present embodiment with
Based on previous embodiment, improved and optimizated.As shown in Fig. 2, image partition method described in the present embodiment includes:
In step S210, image to be split is obtained.
In step S220, by the multi-class identification model of the image input training in advance to be split, according to described more
The output result information of the first object layer of classification identification model obtains first category probability matrix and according to described multi-class
The output result information of second destination layer of identification model obtains second category probability matrix.
Wherein, the second category belongs to the subclass of the first category, the first category probability matrix, described
Two category probability matrix and the size of the image to be split are all the same, the element representation of the first category probability matrix
Corresponding position pixel belongs to the probability value of first category, the list of elements of the second category probability matrix in the image to be split
Show that corresponding position pixel in the image to be split belongs to the probability value of second category.
Above-mentioned steps S210-S220 is identical as step S110-S120 in a upper embodiment, and the present embodiment is not gone to live in the household of one's in-laws on getting married herein
It states.
In step S230, inverse transformation is carried out to the first category probability matrix and obtains the original image corresponding the
One category probability matrix, and inverse transformation is carried out to the second category probability matrix and obtains the original image corresponding second
Category probability matrix.
It is the corresponding first category probability matrix of the original image, described original it should be noted that after carrying out inverse transformation
The size of the corresponding second category probability matrix of image and the original image is all the same.
In step S240, according to the corresponding first category probability matrix of the original image and the original image
Corresponding second category probability matrix carries out repair process to the original image.
For example, threshold value can be carried out to the corresponding first category probability matrix of the original image according to the first given threshold
Change obtains the first category threshold matrix of two-value, according to the second given threshold to the corresponding second category probability of the original image
Matrix carries out thresholding and obtains the second category threshold matrix of two-value, according to the first category threshold matrix and/or described the
Two class threshold matrixes carry out repair process to the image to be split.
For example, in the original image, first category threshold matrix indicates that pixel is spot acne mole classification, and second category
Threshold matrix indicates that pixel is not that the pixel of the mole classification (i.e. classical mole) with setting feature carries out repair process.
The technical solution of the present embodiment is on the basis of a upper embodiment, further to the first category probability matrix
Carry out inverse transformation and obtain the corresponding first category probability matrix of the original image, and to the second category probability matrix into
Row inverse transformation obtains the corresponding second category probability matrix of the original image, according to the corresponding first category of the original image
Probability matrix and the corresponding second category probability matrix of the original image carry out repair process, energy to the original image
It is enough to avoid image fault when carrying out repair process to image.
Fig. 3 is a kind of flow diagram of the training method for multi-class identification model that the embodiment of the present disclosure provides, this reality
Apply multi-class identification model described in example include first category identification submodel and second category identification submodel, as shown in figure 3,
The training method of multi-class identification model described in the present embodiment includes:
In step s310, training sample set is obtained.
Wherein, training sample includes sample image and the mark class for indicating the generic of each pixel in sample image
Other probability matrix, the generic include background classes, first category and second category, wherein the second category belongs to
The subclass of the first category.
In step s 320, submodel is identified according to training sample set training first category, comprising:
It is that the element value of background classes resets to 0 by element value in the corresponding mark category probability matrix of each sample image, member
Plain value resets to 1 for the element value of first category or second category, obtains the corresponding mark first category of each sample image
Probability matrix.
Determine that the first category of initialization identifies submodel, wherein the first category identification submodel of the initialization includes
The first object layer for belonging to the probability of first category for exporting each pixel in target image；It, will using the method for machine learning
Sample image in the training sample set identifies the input of submodel as the first category of initialization, by the sample with input
Desired output of the corresponding mark first category probability matrix of this image as the first category identification submodel of initialization, training
Obtain the first category identification submodel；
In step S330, submodel is identified according to training sample set training second category, comprising:
It is that the element value of background classes resets to setting by element value in the corresponding mark category probability matrix of each sample image
Value, element value are that the element value of first category resets to 0, and element value is that the element value of second category resets to 1, is obtained described each
The corresponding mark second category probability matrix of sample image, wherein the setting value is greater than 1, such as setting value is set as 255.
Determine that the second category of initialization identifies submodel, wherein the second category identification submodel of the initialization includes
The second destination layer for belonging to the probability of second category for exporting each pixel in target image；It, will using the method for machine learning
Sample image in the training sample set identifies the input of submodel as the second category of initialization, by the sample with input
Desired output of the corresponding mark second category probability matrix of this image as the second category identification submodel of initialization, training
Obtain the second category identification submodel.
Further, in the mark category probability matrix, it can indicate that generic is background classes in sample image with 0
Pixel, indicate that generic in sample image is the pixel of first category with 1, indicate that generic is in sample image with 2
The pixel of second category.
The technical solution of the present embodiment discloses a kind of training method of multi-class identification model, by will in sample set it is each
Element value is that the element value of background classes resets to 0 in the corresponding mark category probability matrix of sample image, and element value is the first kind
Not or the element value of second category resets to 1, obtains the corresponding mark first category probability matrix of each sample image；It utilizes
The method of machine learning, using the sample image in the training sample set as the first category of initialization identification submodel
Input identifies submodule using mark first category probability matrix corresponding with the sample image of input as the first category of initialization
The desired output of type, training obtain the first category identification submodel.By the corresponding mark class probability square of each sample image
Element value is that the element value of background classes resets to setting value in battle array, and element value is that the element value of first category resets to 0, element value
1 is reset to for the element value of second category, the corresponding mark second category probability matrix of each sample image is obtained, utilizes machine
The method of device study, using the sample image in the training sample set as the defeated of the second category of initialization identification submodel
Enter, identifies submodel using mark second category probability matrix corresponding with the sample image of input as the second category of initialization
Desired output, training obtains second category identification submodel, with for when performing image segmentation, by image to be split
The multi-class identification model of input training in advance identifies that the output result information of submodel obtains first category according to first category
Probability matrix and according to second category identify submodel output result information obtain second category probability matrix, with quick
Three Classification Semantics segmentations are carried out to image, can be improved image segmentation efficiency.
Fig. 4 shows a kind of structural schematic diagram of image segmentation device of embodiment of the present disclosure offer, as shown in figure 4, this
Image segmentation device described in embodiment includes image acquisition unit 410 and classification recognition unit 420 to be split.
The image acquisition unit to be split 410 is configured to obtain image to be split.
The classification recognition unit 420 is configured to the multi-class of the image input training in advance to be split
Identification model obtains first category probability square according to the output result information of the first object layer of the multi-class identification model
Battle array and second category probability matrix is obtained according to the output result information of the second destination layer of the multi-class identification model.
Wherein, the second category belongs to the subclass of the first category, the first category probability matrix, described
Two category probability matrix and the size of the image to be split are all the same, the element representation of the first category probability matrix
Corresponding position pixel belongs to the probability value of first category, the list of elements of the second category probability matrix in the image to be split
Show that corresponding position pixel in the image to be split belongs to the probability value of second category.
Further, the image to be split is facial image；
The first category is spot acne mole classification, and the second category is the mole classification with setting feature.
Further, the image acquisition unit to be split 410 is configured to obtain the original graph including face
Picture；Facial image to be processed in the original image is obtained as the image to be split.
Further, the image acquisition unit to be split 410 is configured to obtain the collected photograph of camera
The photo is cached in buffer area as the original image including face by piece.
Image provided by embodiment of the present disclosure embodiment of the method point can be performed in image segmentation device provided in this embodiment
Segmentation method has the corresponding functional module of execution method and beneficial effect.
Fig. 5 shows the structural schematic diagram of another image segmentation device of embodiment of the present disclosure offer, as shown in figure 5,
Image segmentation device described in the present embodiment includes image acquisition unit 510 to be split, classification recognition unit 520, inverse transformation list
Member 530 and repair process unit 540.
The image acquisition unit to be split 510 is configured to obtain image to be split；
The classification recognition unit 520 is configured to the multi-class of the image input training in advance to be split
Identification model obtains first category probability square according to the output result information of the first object layer of the multi-class identification model
Battle array and second category probability matrix is obtained according to the output result information of the second destination layer of the multi-class identification model.
Wherein, the second category belongs to the subclass of the first category, the first category probability matrix, described
Two category probability matrix and the size of the image to be split are all the same, the element representation of the first category probability matrix
Corresponding position pixel belongs to the probability value of first category, the list of elements of the second category probability matrix in the image to be split
Show that corresponding position pixel in the image to be split belongs to the probability value of second category.
The inverse transformation unit 530 is configured to obtain institute to first category probability matrix progress inverse transformation
The corresponding first category probability matrix of original image is stated, and second category probability matrix progress inverse transformation is obtained described
The corresponding second category probability matrix of original image, wherein the corresponding first category probability matrix of the original image, the original
The size of the corresponding second category probability matrix of beginning image and the original image is all the same；
The repair process unit 540 is configured to according to the corresponding first category probability square of the original image
Battle array and the corresponding second category probability matrix of the original image carry out repair process to the original image.
Further, the image to be split is facial image；The first category is spot acne mole classification, second class
It Wei not be with the mole classification of setting feature.
Further, the image acquisition unit to be split 510 is configured to: obtaining the original graph including face
Picture；Facial image to be processed in the original image is obtained as the image to be split.
Further, the image acquisition unit to be split 510 is configured to obtain the collected photograph of camera
The photo is cached in buffer area as the original image including face by piece.
Further, the repair process unit 540 includes first threshold beggar unit (being not shown in Fig. 5), the second threshold
Value subelement (being not shown in Fig. 5) and reparation subelement (being not shown in Fig. 5).
The first threshold beggar unit is configured to corresponding to the original image according to the first given threshold
First category probability matrix carries out thresholding and obtains the first category threshold matrix of two-value；And/or
The second threshold beggar unit is configured to corresponding to the original image according to the second given threshold
Second category probability matrix carries out thresholding and obtains the second category threshold matrix of two-value；
The reparation subelement is configured to according to the first category threshold matrix and/or the second category
Threshold matrix carries out repair process to the image to be split.
Further, the reparation subelement is configured to in the original image, first category threshold matrix
Expression pixel is spot acne mole classification, and second category threshold matrix indicates that pixel is not the pixel of the mole classification with setting feature
Carry out repair process.
Image segmentation side provided by disclosed method embodiment can be performed in image segmentation device provided in this embodiment
Method has the corresponding functional module of execution method and beneficial effect.
Fig. 6 is a kind of structural schematic diagram of the training device for multi-class identification model that the embodiment of the present disclosure provides, such as Fig. 6
Shown, multi-class identification model described in the present embodiment includes that first category identification submodel and second category identify submodel,
The training device of the multi-class identification model includes sample acquisition module 610, first category identification submodel training module 620
Submodel training module 630 is identified with second category.
Wherein the sample acquisition module 610 is configured to obtain training sample set, wherein training sample packet
Include sample image and the mark category probability matrix for indicating the generic of each pixel in sample image, the generic
Including background classes, first category and second category, wherein the second category belongs to the subclass of the first category；
The first category identification submodel training module 620 includes the first mark submodule 621 and the first training submodule
Block 622 is configurable for the first category identification submodel according to the training sample set and the training of above-mentioned submodule
The first category identifies submodel.
The first mark submodule 621 is configured to the corresponding mark category probability matrix of each sample image
Middle element value is that the element value of background classes resets to 0, and element value resets to 1 for the element value of first category or second category, obtains
To the corresponding mark first category probability matrix of each sample image；
The first training submodule 622 is configured to determine the first category identification submodel of initialization, wherein
The first category identification submodel of the initialization includes the probability for belonging to first category for exporting each pixel in target image
First object layer；Using the device of machine learning, using the sample image in the training sample set as the of initialization
One classification identifies the input of submodel, using mark first category probability matrix corresponding with the sample image of input as initialization
First category identification submodel desired output, training obtains first category identification submodel；
Second category identifies that submodel training module 630 includes the second mark submodule 631 and the second training submodule
632, the second category identification submodel is configurable for according to the training sample set and above-mentioned submodule training institute
State second category identification submodel.
The second mark submodule 631 is configured to the corresponding mark category probability matrix of each sample image
Middle element value is that the element value of background classes resets to setting value, and element value is that the element value of first category resets to 0, and element value is
The element value of second category resets to 1, obtains the corresponding mark second category probability matrix of each sample image, wherein described
Setting value is greater than 1；
The second training submodule 632 is configured to determine the second category identification submodel of initialization, wherein
The second category identification submodel of the initialization includes the probability for belonging to second category for exporting each pixel in target image
The second destination layer；Using the device of machine learning, using the sample image in the training sample set as the of initialization
Two classifications identify the input of submodel, using mark second category probability matrix corresponding with the sample image of input as initialization
Second category identification submodel desired output, training obtains second category identification submodel.
Further, in the mark category probability matrix, indicate that generic in sample image is the picture of background classes with 0
Element indicates that generic in sample image is the pixel of first category with 1, and use 2 indicates that generic is second in sample image
The pixel of classification.
Further, the setting value is 255.
The training device of multi-class identification model provided in this embodiment can be performed provided by embodiments of the present disclosure
The training method of multi-class identification model has the corresponding functional module of execution method and beneficial effect.
Below with reference to Fig. 7, it illustrates the structural representations for the electronic equipment 700 for being suitable for being used to realize the embodiment of the present disclosure
Figure.Terminal device in the embodiment of the present disclosure can include but is not limited to such as mobile phone, laptop, digital broadcasting and connect
Receive device, PDA (personal digital assistant), PAD (tablet computer), PMP (portable media player), car-mounted terminal (such as vehicle
Carry navigation terminal) etc. mobile terminal and such as number TV, desktop computer etc. fixed terminal.Electricity shown in Fig. 7
Sub- equipment is only an example, should not function to the embodiment of the present disclosure and use scope bring any restrictions.
As shown in fig. 7, electronic equipment 700 may include processing unit (such as central processing unit, graphics processor etc.)
701, random access can be loaded into according to the program being stored in read-only memory (ROM) 702 or from storage device 708
Program in memory (RAM) 703 and execute various movements appropriate and processing.In RAM 703, it is also stored with electronic equipment
Various programs and data needed for 700 operations.Processing unit 701, ROM 702 and RAM 703 pass through the phase each other of bus 704
Even.Input/output (I/O) interface 705 is also connected to bus 704.
In general, following device can connect to I/O interface 705: including such as touch screen, touch tablet, keyboard, mouse, taking the photograph
As the input unit 706 of head, microphone, accelerometer, gyroscope etc.；Including such as liquid crystal display (LCD), loudspeaker, vibration
The output device 707 of dynamic device etc.；Storage device 708 including such as tape, hard disk etc.；And communication device 709.Communication device
709, which can permit electronic equipment 700, is wirelessly or non-wirelessly communicated with other equipment to exchange data.Although Fig. 7 shows tool
There is the electronic equipment 700 of various devices, it should be understood that being not required for implementing or having all devices shown.It can be with
Alternatively implement or have more or fewer devices.
Particularly, it according to the embodiment of the embodiment of the present disclosure, may be implemented as above with reference to the process of flow chart description
Computer software programs.For example, the embodiment of the embodiment of the present disclosure includes a kind of computer program product comprising be carried on meter
Computer program on calculation machine readable medium, the computer program include the program generation for method shown in execution flow chart
Code.In such embodiments, which can be downloaded and installed from network by communication device 709, or
It is mounted from storage device 708, or is mounted from ROM 702.When the computer program is executed by processing unit 701, hold
The above-mentioned function of being limited in the method for the row embodiment of the present disclosure.
It should be noted that the above-mentioned computer-readable medium of the embodiment of the present disclosure can be computer-readable signal media
Or computer readable storage medium either the two any combination.Computer readable storage medium for example can be with
System, device or the device of --- but being not limited to --- electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, or it is any more than
Combination.The more specific example of computer readable storage medium can include but is not limited to: have one or more conducting wires
Electrical connection, portable computer diskette, hard disk, random access storage device (RAM), read-only memory (ROM), erasable type are programmable
Read-only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic are deposited
Memory device or above-mentioned any appropriate combination.In the embodiments of the present disclosure, computer readable storage medium can be any
Include or the tangible medium of storage program, the program can be commanded execution system, device or device and use or tie with it
It closes and uses.And in the embodiments of the present disclosure, computer-readable signal media may include in a base band or as carrier wave one
Divide the data-signal propagated, wherein carrying computer-readable program code.The data-signal of this propagation can use more
Kind form, including but not limited to electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media is also
It can be any computer-readable medium other than computer readable storage medium, which can send out
It send, propagate or transmits for by the use of instruction execution system, device or device or program in connection.It calculates
The program code for including on machine readable medium can transmit with any suitable medium, including but not limited to: electric wire, optical cable, RF
(radio frequency) etc. or above-mentioned any appropriate combination.
Above-mentioned computer-readable medium can be included in above-mentioned electronic equipment；It is also possible to individualism, and not
It is fitted into the electronic equipment.
Above-mentioned computer-readable medium carries one or more program, when said one or multiple programs are by the electricity
When sub- equipment executes, so that the electronic equipment:
Obtain image to be split；
By the multi-class identification model of the image input training in advance to be split, according to the multi-class identification model
The output result information of first object layer obtains first category probability matrix and according to the second of the multi-class identification model
The output result information of destination layer obtains second category probability matrix, wherein the second category belongs to the first category
The size of subclass, the first category probability matrix, the second category probability matrix and the image to be split is homogeneous
Together, corresponding position pixel belongs to the general of first category in image to be split described in the element representation of the first category probability matrix
Rate value, corresponding position pixel belongs to second category in image to be split described in the element representation of the second category probability matrix
Probability value.
The operation for executing the embodiment of the present disclosure can be write with one or more programming languages or combinations thereof
Computer program code, above procedure design language include object oriented program language-such as Java,
Smalltalk, C++ further include conventional procedural programming language-such as " C " language or similar program design language
Speech.Program code can be executed fully on the user computer, partly be executed on the user computer, as an independence
Software package execute, part on the user computer part execute on the remote computer or completely in remote computer or
It is executed on server.In situations involving remote computers, remote computer can pass through the network of any kind --- packet
It includes local area network (LAN) or wide area network (WAN)-is connected to subscriber computer, or, it may be connected to outer computer (such as benefit
It is connected with ISP by internet).
Flow chart and block diagram in attached drawing illustrate system, method and meter according to the various embodiments of the embodiment of the present disclosure
The architecture, function and operation in the cards of calculation machine program product.In this regard, each box in flowchart or block diagram
Can represent a part of a module, program segment or code, a part of the module, program segment or code include one or
Multiple executable instructions for implementing the specified logical function.It should also be noted that in some implementations as replacements, box
Middle marked function can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated
It can actually be basically executed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.
It is also noted that the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart, it can
To be realized with the dedicated hardware based system for executing defined functions or operations, or with specialized hardware and can calculate
The combination of machine instruction is realized.
Being described in unit involved in the embodiment of the present disclosure can be realized by way of software, can also be by hard
The mode of part is realized.Wherein, the title of unit does not constitute the restriction to the unit itself under certain conditions, for example, the
One acquiring unit is also described as " obtaining the unit of at least two internet protocol addresses ".
According to one or more other embodiments of the present disclosure, in described image dividing method, the image to be split is face
Image；The first category is spot acne mole classification, and the second category is the mole classification with setting feature.
According to one or more other embodiments of the present disclosure, in described image dividing method, the operation of image to be split is obtained
Include:
Obtain the original image including face；
Facial image to be processed in the original image is obtained as the image to be split.
According to one or more other embodiments of the present disclosure, in described image dividing method, obtain include face original graph
The operation of picture includes: to obtain the collected photo of camera, and the photo is cached in buffer area as described including face
Original image.
According to one or more other embodiments of the present disclosure, in described image dividing method, the image to be split is being obtained
After corresponding first category probability matrix and second category probability matrix further include:
Inverse transformation is carried out to the first category probability matrix and obtains the corresponding first category probability square of the original image
Battle array, and inverse transformation is carried out to the second category probability matrix and obtains the corresponding second category probability square of the original image
Battle array, wherein the corresponding first category probability matrix of the original image, the corresponding second category probability matrix of the original image,
And the size of the original image is all the same；
According to the corresponding first category probability matrix of the original image and the corresponding second category of the original image
Probability matrix carries out repair process to the original image.
It is corresponding according to the original image in described image dividing method according to one or more other embodiments of the present disclosure
First category probability matrix and the original image corresponding second category probability matrix the original image is repaired
The operation handled again includes:
Thresholding is carried out to the corresponding first category probability matrix of the original image according to the first given threshold and obtains two
The first category threshold matrix of value；And/or
Thresholding is carried out to the corresponding second category probability matrix of the original image according to the second given threshold and obtains two
The second category threshold matrix of value；
The image to be split is carried out according to the first category threshold matrix and/or the second category threshold matrix
Repair process.
According to one or more other embodiments of the present disclosure, in described image dividing method, according to the first category threshold value
Matrix and/or the second category threshold matrix include: to the operation of the image progress repair process to be split
To in the original image, first category threshold matrix indicates that pixel is spot acne mole classification, and second category threshold value
Matrix indicates that pixel is not that the pixel of the mole classification with setting feature carries out repair process.
According to one or more other embodiments of the present disclosure, in described image dividing method, the multi-class identification model packet
First category identification submodel and second category identification submodel are included, training obtains as follows:
Obtain training sample set, wherein training sample includes sample image and for indicating each pixel in sample image
Generic mark category probability matrix, the generic includes background classes, first category and second category,
In, the second category belongs to the subclass of the first category；
Submodel is identified according to the training sample set and the following steps training first category:
It is that the element value of background classes resets to 0 by element value in the corresponding mark category probability matrix of each sample image, member
Plain value resets to 1 for the element value of first category or second category, obtains the corresponding mark first category of each sample image
Probability matrix；
Determine that the first category of initialization identifies submodel, wherein the first category identification submodel of the initialization includes
The first object layer for belonging to the probability of first category for exporting each pixel in target image；It, will using the method for machine learning
Sample image in the training sample set identifies the input of submodel as the first category of initialization, by the sample with input
Desired output of the corresponding mark first category probability matrix of this image as the first category identification submodel of initialization, training
Obtain the first category identification submodel；
Submodel is identified according to the training sample set and the following steps training second category:
It is that the element value of background classes resets to setting by element value in the corresponding mark category probability matrix of each sample image
Value, element value are that the element value of first category resets to 0, and element value is that the element value of second category resets to 1, is obtained described each
The corresponding mark second category probability matrix of sample image, wherein the setting value is greater than 1；
Determine that the second category of initialization identifies submodel, wherein the second category identification submodel of the initialization includes
The second destination layer for belonging to the probability of second category for exporting each pixel in target image；It, will using the method for machine learning
Sample image in the training sample set identifies the input of submodel as the second category of initialization, by the sample with input
Desired output of the corresponding mark second category probability matrix of this image as the second category identification submodel of initialization, training
Obtain the second category identification submodel.
According to one or more other embodiments of the present disclosure, in described image dividing method, the mark category probability matrix
In, it indicates that generic in sample image is the pixel of background classes with 0, indicates that generic is the first kind in sample image with 1
Other pixel indicates that generic in sample image is the pixel of second category with 2.
According to one or more other embodiments of the present disclosure, in described image dividing method, the setting value is 255.
According to one or more other embodiments of the present disclosure, in described image segmenting device, the image to be split is face
Image；
The first category is spot acne mole classification, and the second category is the mole classification with setting feature.
According to one or more other embodiments of the present disclosure, in described image segmenting device, the image to be split obtains single
Member is used for:
Obtain the original image including face；
Facial image to be processed in the original image is obtained as the image to be split.
According to one or more other embodiments of the present disclosure, in described image segmenting device, the image to be split obtains single
Member is used for: obtaining the collected photo of camera, the photo is cached in buffer area as described original including face
Image.
According to one or more other embodiments of the present disclosure, in described image segmenting device, described device further includes inverse transformation
Unit and repair process unit；
The inverse transformation unit is used for, and is obtaining the corresponding first category probability matrix of the image to be split and the second class
After other probability matrix, inverse transformation is carried out to the first category probability matrix and obtains the corresponding first category of the original image
Probability matrix, and inverse transformation is carried out to the second category probability matrix to obtain the corresponding second category of the original image general
Rate matrix, wherein the corresponding first category probability matrix of the original image, the corresponding second category probability of the original image
The size of matrix and the original image is all the same；
The repair process unit is used for, according to the corresponding first category probability matrix of the original image and described
The corresponding second category probability matrix of original image carries out repair process to the original image.
According to one or more other embodiments of the present disclosure, in described image segmenting device, the repair process unit includes
First threshold beggar unit, second threshold beggar unit and reparation subelement；
The first threshold beggar unit is used for, according to the first given threshold to the corresponding first category of the original image
Probability matrix carries out thresholding and obtains the first category threshold matrix of two-value；And/or
The second threshold beggar unit is used for, according to the second given threshold to the corresponding second category of the original image
Probability matrix carries out thresholding and obtains the second category threshold matrix of two-value；
The reparation subelement is used for, according to the first category threshold matrix and/or the second category threshold matrix
Repair process is carried out to the image to be split.
According to one or more other embodiments of the present disclosure, in described image segmenting device, the reparation subelement is used for:
To in the original image, first category threshold matrix indicates that pixel is spot acne mole classification, and second category threshold value
Matrix indicates that pixel is not that the pixel of the mole classification with setting feature carries out repair process.
According to one or more other embodiments of the present disclosure, in described image segmenting device, institute in the classification recognition unit
Stating multi-class identification model includes first category identification submodel and second category identification submodel, the multi-class identification model
It is obtained by following module training:
Sample acquisition module, for obtaining training sample set, wherein training sample includes sample image and for indicating
The mark category probability matrix of the generic of each pixel in sample image, the generic include background classes, first category,
And second category, wherein the second category belongs to the subclass of the first category；
First category identifies submodel training module, for first category identification submodel according to the training sample
Set and the training of following submodule obtain:
First mark submodule, for being background classes by element value in the corresponding mark category probability matrix of each sample image
Element value reset to 0, element value is that the element value of first category or second category resets to 1, obtains each sample image
Corresponding mark first category probability matrix；
First training submodule, for determining that the first category of initialization identifies submodel, wherein the of the initialization
One classification identifies that submodel includes the first object layer for belonging to the probability of first category for exporting each pixel in target image；Benefit
With the device of machine learning, submodel is identified using the sample image in the training sample set as the first category of initialization
Input, mark first category probability matrix corresponding with the sample image of input is sub as the identification of the first category of initialization
The desired output of model, training obtain the first category identification submodel；
Second category identifies submodel training module, for second category identification submodel according to the training sample
Set and the training of following submodule obtain:
Second mark submodule, for being background classes by element value in the corresponding mark category probability matrix of each sample image
Element value reset to setting value, element value is that the element value of first category resets to 0, and element value is the element value of second category
1 is reset to, the corresponding mark second category probability matrix of each sample image is obtained, wherein the setting value is greater than 1；
Second training submodule, for determining that the second category of initialization identifies submodel, wherein the of the initialization
Two classifications identify that submodel includes the second destination layer for belonging to the probability of second category for exporting each pixel in target image；Benefit
With the device of machine learning, submodel is identified using the sample image in the training sample set as the second category of initialization
Input, mark second category probability matrix corresponding with the sample image of input is sub as the identification of the second category of initialization
The desired output of model, training obtain the second category identification submodel.
According to one or more other embodiments of the present disclosure, in described image segmenting device, the mark category probability matrix
In, it indicates that generic in sample image is the pixel of background classes with 0, indicates that generic is the first kind in sample image with 1
Other pixel indicates that generic in sample image is the pixel of second category with 2.
According to one or more other embodiments of the present disclosure, in described image segmenting device, the setting value is 255.
Above description is only the preferred embodiment of the embodiment of the present disclosure and the explanation to institute's application technology principle.This field
It will be appreciated by the skilled person that the open scope involved in the embodiment of the present disclosure, however it is not limited to the specific group of above-mentioned technical characteristic
Technical solution made of conjunction, while should also cover in the case where not departing from design disclosed above, by above-mentioned technical characteristic or its
Equivalent feature carries out any combination and other technical solutions for being formed.Such as disclosed in features described above and the embodiment of the present disclosure
(but being not limited to) have the technical characteristic of similar functions replaced mutually and the technical solution that is formed.
Claims (13)
1. a kind of image partition method characterized by comprising
Obtain image to be split；
By the multi-class identification model of the image input training in advance to be split, according to the first of the multi-class identification model
The output result information of destination layer obtains first category probability matrix and the second target according to the multi-class identification model
The output result information of layer obtains second category probability matrix, wherein the second category belongs to the subclass of the first category
Not, the size of the first category probability matrix, the second category probability matrix and the image to be split is all the same,
Corresponding position pixel belongs to the probability of first category in image to be split described in the element representation of the first category probability matrix
Value, corresponding position pixel belongs to the general of second category in image to be split described in the element representation of the second category probability matrix
Rate value.
2. the method according to claim 1, wherein the image to be split is facial image；
The first category is spot acne mole classification, and the second category is the mole classification with setting feature.
3. according to the method described in claim 2, it is characterized in that, the operation for obtaining image to be split includes:
Obtain the original image including face；
Facial image to be processed in the original image is obtained as the image to be split.
4. according to the method described in claim 3, it is characterized in that, the operation for obtaining the original image including face includes:
The collected photo of camera is obtained, the photo is cached in buffer area as the original graph including face
Picture.
5. according to the method described in claim 3, it is characterized in that, general obtaining the corresponding first category of the image to be split
After rate matrix and second category probability matrix further include:
Inverse transformation is carried out to the first category probability matrix and obtains the corresponding first category probability matrix of the original image, with
And inverse transformation is carried out to the second category probability matrix and obtains the corresponding second category probability matrix of the original image, wherein
The corresponding first category probability matrix of the original image, the corresponding second category probability matrix of the original image, Yi Jisuo
The size for stating original image is all the same；
According to the corresponding first category probability matrix of the original image and the corresponding second category probability of the original image
Matrix carries out repair process to the original image.
6. according to the method described in claim 5, it is characterized in that, according to the corresponding first category probability square of the original image
Battle array and the corresponding second category probability matrix of the original image carry out the operation packet of repair process to the original image
It includes:
Thresholding is carried out to the corresponding first category probability matrix of the original image according to the first given threshold and obtains two-value
First category threshold matrix；And/or
Thresholding is carried out to the corresponding second category probability matrix of the original image according to the second given threshold and obtains two-value
Second category threshold matrix；
The image to be split is repaired according to the first category threshold matrix and/or the second category threshold matrix
Processing.
7. according to the method described in claim 6, it is characterized in that, according to the first category threshold matrix and/or described
The operation that two class threshold matrixes carry out repair process to the image to be split includes:
To in the original image, first category threshold matrix indicates that pixel is spot acne mole classification, and second category threshold matrix
Indicate that pixel is not that the pixel of the mole classification with setting feature carries out repair process.
8. method described in one of -7 according to claim 1, which is characterized in that the multi-class identification model includes first category
Identify that submodel and second category identify submodel, training obtains the multi-class identification model as follows:
Obtain training sample set, wherein training sample includes sample image and the institute for indicating each pixel in sample image
Belong to the mark category probability matrix of classification, the generic includes background classes, first category and second category, wherein institute
State the subclass that second category belongs to the first category；
Submodel is identified according to the training sample set and the following steps training first category:
It is that the element value of background classes resets to 0 by element value in the corresponding mark category probability matrix of each sample image, element value
Element value for first category or second category resets to 1, obtains the corresponding mark first category probability of each sample image
Matrix；
Determine that the first category of initialization identifies submodel, wherein the first category identification submodel of the initialization includes being used for
Each pixel belongs to the first object layer of the probability of first category in output target image；It, will be described using the method for machine learning
Sample image in training sample set identifies the input of submodel as the first category of initialization, by the sample graph with input
Desired output as corresponding mark first category probability matrix as the first category identification submodel of initialization, training obtain
The first category identifies submodel；
Submodel is identified according to the training sample set and the following steps training second category:
It is that the element value of background classes resets to setting value by element value in the corresponding mark category probability matrix of each sample image, member
Plain value is that the element value of first category resets to 0, and element value is that the element value of second category resets to 1, obtains each sample
The corresponding mark second category probability matrix of image, wherein the setting value is greater than 1；
Determine that the second category of initialization identifies submodel, wherein the second category identification submodel of the initialization includes being used for
Each pixel belongs to the second destination layer of the probability of second category in output target image；It, will be described using the method for machine learning
Sample image in training sample set identifies the input of submodel as the second category of initialization, by the sample graph with input
Desired output as corresponding mark second category probability matrix as the second category identification submodel of initialization, training obtain
The second category identifies submodel.
9. the method according to claim 1, wherein use 0 indicates sample graph in the mark category probability matrix
Generic is the pixel of background classes as in, indicates that generic in sample image is the pixel of first category with 1, use 2 indicates
Generic is the pixel of second category in sample image.
10. the method according to claim 1, wherein the setting value is 255.
11. a kind of image segmentation device characterized by comprising
Image acquisition unit to be split, for obtaining image to be split；
Classification recognition unit, the multi-class identification model for training the image input to be split in advance, according to described more
The output result information of the first object layer of classification identification model obtains first category probability matrix and according to described multi-class
The output result information of second destination layer of identification model obtains second category probability matrix, wherein the second category belongs to
The subclass of the first category, the first category probability matrix, the second category probability matrix and described to be split
The size of image is all the same, corresponding position pixel category in image to be split described in the element representation of the first category probability matrix
In the probability value of first category, corresponding position pixel in image to be split described in the element representation of the second category probability matrix
Belong to the probability value of second category.
12. a kind of electronic equipment characterized by comprising
One or more processors；
Memory, for storing one or more programs；
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
Now such as the instruction of any one of claim 1-10 the method.
13. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program quilt
It is realized when processor executes such as the step of any one of claim 1-10 the method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910621621.4A CN110349161B (en) | 2019-07-10 | 2019-07-10 | Image segmentation method, image segmentation device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910621621.4A CN110349161B (en) | 2019-07-10 | 2019-07-10 | Image segmentation method, image segmentation device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110349161A true CN110349161A (en) | 2019-10-18 |
| CN110349161B CN110349161B (en) | 2021-11-23 |
Family
ID=68174765
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910621621.4A Active CN110349161B (en) | 2019-07-10 | 2019-07-10 | Image segmentation method, image segmentation device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110349161B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110837789A (en) * | 2019-10-31 | 2020-02-25 | 北京奇艺世纪科技有限公司 | Method and device for detecting object, electronic equipment and medium |
| CN111046944A (en) * | 2019-12-10 | 2020-04-21 | 北京奇艺世纪科技有限公司 | Method and device for determining object class, electronic equipment and storage medium |
| CN111310815A (en) * | 2020-02-07 | 2020-06-19 | 北京字节跳动网络技术有限公司 | Image recognition method and device, electronic equipment and storage medium |
| CN114565759A (en) * | 2022-02-22 | 2022-05-31 | 北京百度网讯科技有限公司 | Image semantic segmentation model optimization method and device, electronic equipment and storage medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150030255A1 (en) * | 2013-07-25 | 2015-01-29 | Canon Kabushiki Kaisha | Method and apparatus for classifying pixels in an input image and image processing system |
| US20160284084A1 (en) * | 2015-03-23 | 2016-09-29 | Ohio State Innovation Foundation | System and method for segmentation and automated measurement of chronic wound images |
| CN107203999A (en) * | 2017-04-28 | 2017-09-26 | 北京航空航天大学 | A kind of skin lens image automatic division method based on full convolutional neural networks |
| CN108010021A (en) * | 2017-11-30 | 2018-05-08 | 上海联影医疗科技有限公司 | A kind of magic magiscan and method |
| CN108256476A (en) * | 2018-01-17 | 2018-07-06 | 百度在线网络技术（北京）有限公司 | For identifying the method and apparatus of fruits and vegetables |
| US10096122B1 (en) * | 2017-03-28 | 2018-10-09 | Amazon Technologies, Inc. | Segmentation of object image data from background image data |
| CN109344752A (en) * | 2018-09-20 | 2019-02-15 | 北京字节跳动网络技术有限公司 | Method and apparatus for handling mouth image |
| CN109784283A (en) * | 2019-01-21 | 2019-05-21 | 陕西师范大学 | Remote sensing image target extraction method based on scene recognition task |
| CN109886986A (en) * | 2019-01-23 | 2019-06-14 | 北京航空航天大学 | A Dermoscopy Image Segmentation Method Based on Multi-branch Convolutional Neural Networks |
-
2019
- 2019-07-10 CN CN201910621621.4A patent/CN110349161B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150030255A1 (en) * | 2013-07-25 | 2015-01-29 | Canon Kabushiki Kaisha | Method and apparatus for classifying pixels in an input image and image processing system |
| US20160284084A1 (en) * | 2015-03-23 | 2016-09-29 | Ohio State Innovation Foundation | System and method for segmentation and automated measurement of chronic wound images |
| US10096122B1 (en) * | 2017-03-28 | 2018-10-09 | Amazon Technologies, Inc. | Segmentation of object image data from background image data |
| CN107203999A (en) * | 2017-04-28 | 2017-09-26 | 北京航空航天大学 | A kind of skin lens image automatic division method based on full convolutional neural networks |
| CN108010021A (en) * | 2017-11-30 | 2018-05-08 | 上海联影医疗科技有限公司 | A kind of magic magiscan and method |
| CN108256476A (en) * | 2018-01-17 | 2018-07-06 | 百度在线网络技术（北京）有限公司 | For identifying the method and apparatus of fruits and vegetables |
| CN109344752A (en) * | 2018-09-20 | 2019-02-15 | 北京字节跳动网络技术有限公司 | Method and apparatus for handling mouth image |
| CN109784283A (en) * | 2019-01-21 | 2019-05-21 | 陕西师范大学 | Remote sensing image target extraction method based on scene recognition task |
| CN109886986A (en) * | 2019-01-23 | 2019-06-14 | 北京航空航天大学 | A Dermoscopy Image Segmentation Method Based on Multi-branch Convolutional Neural Networks |
Non-Patent Citations (3)
| Title |
|---|
| DHARANIBAI,G 等: "Automated Segmentation of Left Ventricle Using Local and Global Intensity Based Active Contour and Dynamic Programming", 《INTERNATIONAL JOURNAL OF AUTOMATION AND COMPUTING》 * |
| 夏良正 主编: "《数字图像处理 修订版》", 30 September 1999 * |
| 姜枫 等: "基于内容的图像分割方法综述", 《软件学报》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110837789A (en) * | 2019-10-31 | 2020-02-25 | 北京奇艺世纪科技有限公司 | Method and device for detecting object, electronic equipment and medium |
| CN111046944A (en) * | 2019-12-10 | 2020-04-21 | 北京奇艺世纪科技有限公司 | Method and device for determining object class, electronic equipment and storage medium |
| CN111310815A (en) * | 2020-02-07 | 2020-06-19 | 北京字节跳动网络技术有限公司 | Image recognition method and device, electronic equipment and storage medium |
| CN114565759A (en) * | 2022-02-22 | 2022-05-31 | 北京百度网讯科技有限公司 | Image semantic segmentation model optimization method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110349161B (en) | 2021-11-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112528977B (en) | Target detection method, target detection device, electronic equipment and storage medium | |
| CN111368685B (en) | Method and device for identifying key points, readable medium and electronic equipment | |
| CN110349161A (en) | Image partition method, device, electronic equipment and storage medium | |
| CN109829432B (en) | Method and apparatus for generating information | |
| US11164004B2 (en) | Keyframe scheduling method and apparatus, electronic device, program and medium | |
| US11100325B2 (en) | Photographic results by composition analysis using deep learning neural networks | |
| CN110059623B (en) | Method and apparatus for generating information | |
| US11025500B2 (en) | Provisioning infrastructure from visual diagrams | |
| CN113378855A (en) | Method for processing multitask, related device and computer program product | |
| CN110211195B (en) | Method, device, electronic equipment and computer-readable storage medium for generating image set | |
| CN110633717A (en) | Training method and device for target detection model | |
| CN110427915A (en) | Method and apparatus for output information | |
| CN113657518A (en) | Training method, target image detection method, device, electronic device, and medium | |
| WO2021164328A1 (en) | Image generation method, device, and storage medium | |
| CN110969641A (en) | Image processing method and device | |
| CN110110696B (en) | Method and apparatus for processing information | |
| CN113837194B (en) | Image processing method, image processing apparatus, electronic device, and storage medium | |
| CN112784102B (en) | Video retrieval method and device and electronic equipment | |
| CN109829431A (en) | Method and apparatus for generating information | |
| CN108537149A (en) | Image processing method, device, storage medium and electronic equipment | |
| CN114119990A (en) | Method, apparatus and computer program product for image feature point matching | |
| CN113742485A (en) | Method and device for processing text | |
| CN113627354B (en) | A model training and video processing method, which comprises the following steps, apparatus, device, and storage medium | |
| CN110852250A (en) | Vehicle weight removing method and device based on maximum area method and storage medium | |
| CN111428536A (en) | Training method and device for detection network for detecting article category and position |
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 |