CN110097081A - A kind of construction method and device of training set - Google Patents
A kind of construction method and device of training set Download PDFInfo
- Publication number
- CN110097081A CN110097081A CN201910251169.7A CN201910251169A CN110097081A CN 110097081 A CN110097081 A CN 110097081A CN 201910251169 A CN201910251169 A CN 201910251169A CN 110097081 A CN110097081 A CN 110097081A
- Authority
- CN
- China
- Prior art keywords
- training set
- picture
- object picture
- note
- classification
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2431—Multiple classes
Abstract
The embodiment of the present invention provides the construction method and device of a kind of training set, which comprises obtains the samples pictures for training preset model；The preset model is the primary dcreening operation network model of original image for identification；Classify to all samples pictures, and according to classification host type construct respectively include similar samples pictures training set；Wherein, the classification host type includes that picture, shooting exterior surface is interfered not to include the first object picture of off-note, shoot the second Target Photo that exterior surface includes the off-note, and the off-note includes protruding features and/or designated color feature.Described device executes the above method.The construction method and device of training set provided in an embodiment of the present invention, to for training all samples pictures of primary dcreening operation network model to classify, and training set is constructed according to classification host type respectively, the reasonability that can be improved training set building, so that the model obtained by the training set can more accurately identify picture.
Description
Technical field
The present embodiments relate to image processing technology more particularly to the construction methods and device of a kind of training set.
Background technique
Capsule endoscope have many advantages, such as it is painless, without wound, shooting image contain much information, have wide application value.
The prior art is identified by the original image of capsule endoscope shooting using manual type and divides original image
Class needs to construct model to more accurately and efficiently identify original image, but model usually require before the use into
Row is trained, and the training set in training process is constructed, so that model can more accurately carry out picture recognition, but
It is the building for the training set of the model of original image for identification, there are no effective methods for the prior art.
Therefore, drawbacks described above how is avoided, the reasonability of training set building is improved, so that obtaining by the training set
Model can more accurately identify picture, becoming need solve the problems, such as.
Summary of the invention
In view of the problems of the existing technology, the embodiment of the present invention provides the construction method and device of a kind of training set.
The embodiment of the present invention provides a kind of construction method of training set, comprising:
Obtain the samples pictures for training preset model；The preset model is the first sieve of original image for identification
Network model；
Classify to all samples pictures, and according to classification host type construct respectively include similar samples pictures instruction
Practice collection；Wherein, the classification host type includes that picture, shooting exterior surface is interfered not to include the first object figure of off-note
Piece, shooting exterior surface include the second Target Photo of the off-note, and the off-note includes protruding features and/or refers to
Determine color characteristic.
The embodiment of the present invention provides a kind of construction device of training set, comprising:
Acquiring unit, for obtaining the samples pictures for training preset model；The preset model is former for identification
The primary dcreening operation network model of beginning picture；
Construction unit, for classifying to all samples pictures, and it includes same for being constructed respectively according to classification host type
The training set of class samples pictures；Wherein, the classification host type includes interference picture, shooting exterior surface not comprising off-note
First object picture, shooting exterior surface include the off-note the second Target Photo, the off-note includes convex
Play feature and/or designated color feature.
The embodiment of the present invention provides a kind of electronic equipment, comprising: processor, memory and bus, wherein
The processor and the memory complete mutual communication by the bus；
The memory is stored with the program instruction that can be executed by the processor, and the processor calls described program to refer to
Order is able to carry out following method:
Obtain the samples pictures for training preset model；The preset model is the first sieve of original image for identification
Network model；
Classify to all samples pictures, and according to classification host type construct respectively include similar samples pictures instruction
Practice collection；Wherein, the classification host type includes that picture, shooting exterior surface is interfered not to include the first object figure of off-note
Piece, shooting exterior surface include the second Target Photo of the off-note, and the off-note includes protruding features and/or refers to
Determine color characteristic.
The embodiment of the present invention provides a kind of non-transient computer readable storage medium, comprising:
The non-transient computer readable storage medium stores computer instruction, and the computer instruction makes the computer
Execute following method:
Obtain the samples pictures for training preset model；The preset model is the first sieve of original image for identification
Network model；
Classify to all samples pictures, and according to classification host type construct respectively include similar samples pictures instruction
Practice collection；Wherein, the classification host type includes that picture, shooting exterior surface is interfered not to include the first object figure of off-note
Piece, shooting exterior surface include the second Target Photo of the off-note, and the off-note includes protruding features and/or refers to
Determine color characteristic.
The construction method and device of training set provided in an embodiment of the present invention, to for training primary dcreening operation network model to own
Samples pictures are classified, and construct training set respectively according to classification host type, can be improved the reasonability of training set building, into
And enables and picture is more accurately identified by the model that the training set obtains.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is the construction method embodiment flow chart of training set of the present invention；
Fig. 2 (a)~Fig. 2 (h) is the screenshot of the full exposure image of shooting of the embodiment of the present invention；
Fig. 3 (a)~Fig. 3 (g) is the screenshot of the picture before capsule endoscope entrance of shooting of the embodiment of the present invention；
Fig. 4 (a)~Fig. 4 (g) is the screenshot of the whole figure picture of homogeneous of shooting of the embodiment of the present invention；
Fig. 5 (a)~Fig. 5 (g) is the screenshot of the original image for being attached with covering of shooting of the embodiment of the present invention；
Fig. 6 (a)~Fig. 6 (h) be the embodiment of the present invention shooting include slaking residue object original image screenshot；
Fig. 7 (a)~Fig. 7 (h) is the first object picture partially with change in shape of shooting of the embodiment of the present invention
Screenshot；
Fig. 8 (a)~Fig. 8 (h) is the first object picture partially with tone variations of shooting of the embodiment of the present invention
Screenshot；
Fig. 9 (a)~Fig. 9 (f) is that the overall situation of shooting of the embodiment of the present invention has the first object picture of stomach corner structure
Screenshot；
Figure 10 (a)~Figure 10 (f) is that the overall situation of shooting of the embodiment of the present invention has the first object picture of texture structure
Screenshot；
Figure 11 (a)~Figure 11 (h) is the screenshot of the hole shape structure first object picture of shooting of the embodiment of the present invention；
Figure 12 is the construction device example structure schematic diagram of training set of the present invention；
Figure 13 is electronic equipment entity structure schematic diagram provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
Fig. 1 is the construction method embodiment flow chart of training set of the present invention, as shown in Figure 1, provided in an embodiment of the present invention
A kind of construction method of training set, comprising the following steps:
S101: the samples pictures for training preset model are obtained；The preset model is original image for identification
Primary dcreening operation network model.
Specifically, device obtains the samples pictures for training preset model；The preset model is original for identification
The primary dcreening operation network model of picture.It should be understood that the original image is shot by capsule endoscope, to the work of capsule endoscope
It is explained as follows as process:
Capsule endoscope enters alimentary canal from oral cavity, then naturally drains in vitro from anus.
The battery durable power of capsule endoscope is limited, and effective operation interval is oral cavity, esophagus, Stomach duodenum, small intestine
With large intestine a part.
Each activity of capsule endoscope, which all generates, checks picture and overseas inspection picture in domain.
Check that picture is to a certain section of shooting result carried out of alimentary canal in domain.
Overseas inspection picture is the picture that capsule endoscope photographed in passing other than checking picture in domain.
Whole pictures can automatic identification, be not necessarily to any manpower intervention (including image preprocessing).
Identify image after, by capsule endoscope shoot picture be divided into six major class (125 groups), automatically save in
In 125 Photo folders, wherein six major class can be with are as follows:
First major class: a kind of overseas tag along sort (10 classifications).
Second major class: the overseas tag along sort of two classes (13 classifications).
Third major class: the first object picture classification label (14 classifications) based on partial structurtes feature.
The fourth-largest class: hole shape structure first object picture classification label (8 classifications).
The fifth-largest class: the first object picture classification label (24 classifications) based on global structure feature.
The sixth-largest class: the second Target Photo tag along sort (56 classifications).
It being capable of the gastral different parts such as automatic identification oral cavity, esophagus, Stomach duodenum, small intestine and large intestine.
The quantity for the original image that every capsule endoscope can be shot every time can be 2000~3000, i.e. capsule endoscope
The picture number in pictures got.
It can be exported from hospital information system, original image (the JPG lattice that the capsule endoscope without any processing is shot
Formula).Above-mentioned samples pictures can be the corresponding all pictures of six above-mentioned major class (125 groups).Primary dcreening operation network model can
Think convolutional neural networks inceptionV3, input original image to primary dcreening operation network model, the output result of primary dcreening operation network model
It may include interference picture, interference picture can be understood as being not used to the picture of picture recognition, shooting exterior surface does not include
First object picture, the shooting exterior surface of off-note include the second Target Photo of the off-note.Off-note can
With include protruding features and/or designated color feature, protruding features may include swelling, granular substance protrusion.Designated color is special
Sign may include red, white, be not especially limited.It, can be with when preset model output result includes the second Target Photo
The special marking for being directed to off-note is generated, such as selects off-note with box frame, to indicate that related personnel selects party's circle
Part is carefully checked, i.e., off-note can be used as the fixed reference feature during certain medicals diagnosis on disease, and it is different only to rely only on this
Chang Tezheng is also not enough to be diagnosed to be disease.It should be understood that the input of primary dcreening operation network model is whole original image, and due to
The characteristic of medical image, such as complexity, not being easily distinguishable property, so that the output result of primary dcreening operation network model is not accurate enough.This hair
The training set that bright embodiment is related to constructing is the training set for training the primary dcreening operation network model.
S102: classifying to all samples pictures, and being constructed respectively according to classification host type includes similar sample graph
The training set of piece；Wherein, the classification host type includes that picture, shooting exterior surface is interfered not to include the first mesh of off-note
Mark on a map piece, shooting exterior surface include the off-note the second Target Photo, the off-note include protruding features and/
Or designated color feature.
Specifically, device classifies to all samples pictures, and it includes similar for being constructed respectively according to classification host type
The training set of samples pictures；Wherein, the classification host type includes interference picture, shooting exterior surface not comprising off-note
First object picture, shooting exterior surface include the second Target Photo of the off-note, and the off-note includes protrusion
Feature and/or designated color feature.Interfering the corresponding classification host type of picture is the overseas tag along sort of above-mentioned one kind and two classes
Overseas tag along sort；The corresponding classification host type of first object picture is the above-mentioned first object figure based on partial structurtes feature
Piece tag along sort, hole shape structure first object picture classification label and the first object picture classification mark based on global structure feature
Label.To according to classification host type construct respectively include similar samples pictures training set, be illustrated below: interfere picture pair
The training set for the classification host type answered is A, and all samples pictures are all a kind of overseas tag along sort and two class fields in training set A
The corresponding picture of outer tag along sort, i.e. similar samples pictures in training set A are a kind of overseas tag along sort and overseas point of two classes
The corresponding picture of class label；The training set of the corresponding classification host type of first object picture is B, all samples in training set B
Picture be all based on the first object picture classification label of partial structurtes feature, hole shape structure first object picture classification label and
The corresponding picture of first object picture classification label based on global structure feature, i.e. similar samples pictures in training set B are
First object picture classification label, hole shape structure first object picture classification label based on partial structurtes feature and based on the overall situation
The corresponding picture of first object picture classification label of structure feature；The training set of the corresponding classification host type of second Target Photo
For C, all samples pictures are all the second Target Photos in training set C.
The construction method of training set provided in an embodiment of the present invention, to for training all sample graphs of primary dcreening operation network model
Piece is classified, and constructs training set respectively according to classification host type, can be improved the reasonability of training set building, so that
The model obtained by the training set can more accurately identify picture.
On the basis of the above embodiments, the corresponding classification host type of the interference picture further includes a kind of overseas contingency table
Label and the overseas tag along sort of two classes；Correspondingly, the method also includes:
According to the overseas tag along sort of described one kind and the overseas tag along sort of two classes, construct and the interference picture respectively
The sub- training set of corresponding training set；Wherein, the overseas tag along sort of one kind be shooting defect based on the original image, with
The unrelated shooting position of target site to be detected determines；The overseas tag along sort of two classes is worth based on no medical judgment
Original image, the original image for being attached with covering, include slaking residue object original image determine.
Specifically, device is according to the overseas tag along sort of described one kind and the overseas tag along sort of two classes, respectively building with
The interference picture corresponds to the sub- training set of training set；Wherein, the overseas tag along sort of one kind is based on the original image
Shooting defect, unrelated with target site to be detected shooting position determines；The overseas tag along sort of two classes is based on nothing
Medical judgment value original image, be attached with covering original image, include slaking residue object original image determine
's.To the sub- training set for constructing training set corresponding with the interference picture respectively, it is illustrated below:
Referring to the example above, the sub- training set A1 of training set A is corresponded to according to a kind of overseas tag along sort building interference picture,
It is all the corresponding picture of the overseas tag along sort of one kind in i.e. sub- training set A1；According to the overseas tag along sort building interference picture of two classes
It is all the corresponding picture of the overseas tag along sort of two classes in the sub- training set A2 of corresponding training set A, i.e., sub- training set A2.Original image
Shooting defect may include full exposure image, be not especially limited.Fig. 2 (a)~Fig. 2 (h) is shooting of the embodiment of the present invention
Full exposure image screenshot；The shooting position unrelated with target site to be detected may include before capsule endoscope entrance
Picture is not especially limited.Fig. 3 (a)~Fig. 3 (g) is the figure before capsule endoscope entrance of shooting of the embodiment of the present invention
The screenshot of piece.The original image of no medical judgment value may include the whole figure picture of homogeneous, be described as follows: subject
Surface flat-satin, without significant texture, color is uniform, although shooting quality is very high, since content is excessively single, has lost
Medical judgment is gone to be worth (can not judge the location of reference object, angle, organ carrier, anatomical features etc.).The number of picture
Measuring accounting is about 5.8%, this ratio is very high.This kind of picture is due to losing medical value, although on surface not being rubbish
Picture, i.e. interference picture, but actually distinguish with " rubbish picture " without what, it can ignore completely in subsequent processes,
Fig. 4 (a)~Fig. 4 (g) is the screenshot of the whole figure picture of homogeneous of shooting of the embodiment of the present invention.Fig. 5 (a)~Fig. 5 (g) is this hair
The screenshot of the original image for being attached with covering of bright embodiment shooting, covering may include: lumps suspended matter, bubble population
It with mucous membrane body, is not especially limited, the original image for being attached with covering can be as shown in Fig. 5 (a)~Fig. 5 (g).Fig. 6 (a)~
Fig. 6 (h) be the embodiment of the present invention shooting include slaking residue object original image screenshot, include slaking residue object
Original image can be as shown in Fig. 6 (a)~Fig. 6 (h).
The construction method of training set provided in an embodiment of the present invention, by constructing training set corresponding with interference picture respectively
Sub- training set is further able to improve the reasonability of training set building, so that can by the model that the training set obtains
More accurately identify picture.
On the basis of the above embodiments, the corresponding classification host type of the first object picture further includes based on part knot
The first object picture classification label of structure feature, hole shape structure first object picture classification label, based on global structure feature
First object picture classification label；Correspondingly, the method also includes:
According to first object picture classification label, hole shape structure first object picture classification mark based on partial structurtes feature
Label, the first object picture classification label based on global structure feature construct training corresponding with the first object picture respectively
The sub- training set of collection；Wherein, the first object picture classification label based on partial structurtes feature is based on partially with shape
Shape and/or the first object picture of tone variations are come the first object picture classification based on global structure feature determine, described
Label is determined based on the global first object picture with stomach corner structure and/or texture structure.
Specifically, device is according to first object picture classification label, hole the first mesh of shape structure based on partial structurtes feature
It marks on a map piece tag along sort, the first object picture classification label based on global structure feature, respectively building and the first object
Picture corresponds to the sub- training set of training set；Wherein, the first object picture classification label based on partial structurtes feature is base
In the first object picture partially with shape and/or tone variations come first based on global structure feature determine, described
Target Photo tag along sort is determined based on the global first object picture with stomach corner structure and/or texture structure.
Referring to the example above, first object picture pair is constructed according to the first object picture classification label of partial structurtes feature
The sub- training set B1 of training set B is answered, i.e., is all the first object picture classification label pair of partial structurtes feature in sub- training set B1
The picture answered；Similarly, for hole shape structure first object picture classification label, based on the first object picture of global structure feature
The explanation of tag along sort, repeats no more.Fig. 7 (a)~Fig. 7 (h) is shooting of the embodiment of the present invention partially with change in shape
Screenshot, Fig. 8 (a)~Fig. 8 (h) of first object picture be shooting of the embodiment of the present invention partially with the of tone variations
The screenshot of one Target Photo, the first object picture classification label based on partial structurtes feature can as Fig. 7 (a)~Fig. 7 (h),
Shown in Fig. 8 (a)~Fig. 8 (h).Fig. 9 (a)~Fig. 9 (f) is that the overall situation of shooting of the embodiment of the present invention has the first of stomach corner structure
Screenshot, Figure 10 (a)~Figure 10 (f) of Target Photo are that the overall situation of shooting of the embodiment of the present invention has the first mesh of texture structure
Marking the screenshot of piece, the first object picture classification label based on global structure feature on a map can be such as Fig. 9 (a)~Fig. 9 (f), 10 (a)
Shown in~Figure 10 (f).Figure 11 (a)~Figure 11 (h) is cutting for the hole shape structure first object picture of shooting of the embodiment of the present invention
Figure, hole shape structure first object picture classification label can be as shown in Figure 11 (a)~Figure 11 (h).
The construction method of training set provided in an embodiment of the present invention, by constructing training corresponding with first object picture respectively
The sub- training set of collection is further able to improve the reasonability of training set building, so that the model obtained by the training set
It can more accurately identify picture.
On the basis of the above embodiments, the covering includes:
Lumps suspended matter, bubble population and mucous membrane body.
Specifically, the covering in device includes: lumps suspended matter, bubble population and mucous membrane body.It can refer to above-mentioned
Illustrate, repeats no more.
The construction method of training set provided in an embodiment of the present invention can be more by limiting the particular content of covering
The classification for segmenting training set is further able to improve the reasonability of training set building, so that obtained by the training set
Model can more accurately identify picture.
On the basis of the above embodiments, the target site to be detected is stomach.
Specifically, the target site to be detected in device is stomach.It can refer to above description, repeat no more.
The construction method of training set provided in an embodiment of the present invention, further can be more by the model that the training set obtains
Add accurately identification stomach picture.
On the basis of the above embodiments, the primary dcreening operation network model is convolutional neural networks inceptionV3.
Specifically, the primary dcreening operation network model in device is convolutional neural networks inceptionV3.It can refer to and state
It is bright, it repeats no more.
The construction method of training set provided in an embodiment of the present invention, by the way that primary dcreening operation network model is selected as convolutional neural networks
InceptionV3 facilitates efficiently primary dcreening operation picture.
Figure 12 is the construction device example structure schematic diagram of training set of the present invention, as shown in figure 12, the embodiment of the present invention
Provide a kind of construction device of training set, including acquiring unit 1201 and construction unit 1202, in which:
Acquiring unit 1201 is used to obtain the samples pictures for training preset model；The preset model is for identification
The primary dcreening operation network model of original image；Construction unit 1202 is used to classify to all samples pictures, and according to classification main classes
Type construct respectively include similar samples pictures training set；Wherein, the classification host type includes interference picture, outside subject
Surface does not include the first object picture of off-note, shooting exterior surface includes the second Target Photo of the off-note,
The off-note includes protruding features and/or designated color feature.
Specifically, acquiring unit 1201 is used to obtain the samples pictures for training preset model；The preset model is
The primary dcreening operation network model of original image for identification；Construction unit 1202 is used to classify to all samples pictures, and according to
Classification host type construct respectively include similar samples pictures training set；Wherein, the classification host type include interference picture,
Shoot the second mesh of first object picture, shooting exterior surface comprising the off-note that exterior surface does not include off-note
It marks on a map piece, the off-note includes protruding features and/or designated color feature.
The construction method of training set provided in an embodiment of the present invention, to for training all sample graphs of primary dcreening operation network model
Piece is classified, and constructs training set respectively according to classification host type, can be improved the reasonability of training set building, so that
The model obtained by the training set can more accurately identify picture.
The construction method of training set provided in an embodiment of the present invention specifically can be used for executing above-mentioned each method embodiment
Process flow, details are not described herein for function, is referred to the detailed description of above method embodiment.
Figure 13 is electronic equipment entity structure schematic diagram provided in an embodiment of the present invention, and as shown in figure 13, the electronics is set
Standby includes: processor (processor) 1301, memory (memory) 1302 and bus 1303；
Wherein, the processor 1301, memory 1302 complete mutual communication by bus 1303；
The processor 1301 is used to call the program instruction in the memory 1302, is implemented with executing above-mentioned each method
Method provided by example, for example, classify to all samples pictures, and constructed respectively according to classification host type and include
The training set of similar samples pictures；Wherein, the classification host type includes interference picture, shooting exterior surface not comprising abnormal special
First object picture, the shooting exterior surface of sign include the second Target Photo of the off-note, and the off-note includes
Protruding features and/or designated color feature.
The present embodiment discloses a kind of computer program product, and the computer program product includes being stored in non-transient calculating
Computer program on machine readable storage medium storing program for executing, the computer program include program instruction, when described program instruction is calculated
Machine execute when, computer is able to carry out method provided by above-mentioned each method embodiment, for example, to all samples pictures into
Row classification, and according to classification host type construct respectively include similar samples pictures training set；Wherein, the classification host type
The first object picture of off-note is not included including interference picture, shooting exterior surface, shooting exterior surface includes described different
The second Target Photo of Chang Tezheng, the off-note include protruding features and/or designated color feature.
The present embodiment provides a kind of non-transient computer readable storage medium, the non-transient computer readable storage medium
Computer instruction is stored, the computer instruction makes the computer execute method provided by above-mentioned each method embodiment, example
Such as include: classify to all samples pictures, and according to classification host type construct respectively include similar samples pictures instruction
Practice collection；Wherein, the classification host type includes that picture, shooting exterior surface is interfered not to include the first object figure of off-note
Piece, shooting exterior surface include the second Target Photo of the off-note, and the off-note includes protruding features and/or refers to
Determine color characteristic.
Those of ordinary skill in the art will appreciate that: realize that all or part of the steps of above method embodiment can pass through
The relevant hardware of program instruction is completed, and program above-mentioned can be stored in a computer readable storage medium, the program
When being executed, step including the steps of the foregoing method embodiments is executed；And storage medium above-mentioned includes: ROM, RAM, magnetic disk or light
The various media that can store program code such as disk.
The apparatus embodiments described above are merely exemplary, wherein described, unit can as illustrated by the separation member
It is physically separated with being or may not be, component shown as a unit may or may not be physics list
Member, it can it is in one place, or may be distributed over multiple network units.It can be selected according to the actual needs
In some or all of the modules achieve the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying creativeness
Labour in the case where, it can understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It realizes by means of software and necessary general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be embodied in the form of software products in other words, should
Computer software product may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation
Method described in certain parts of example or embodiment.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features；
And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (9)
1. a kind of construction method of training set characterized by comprising
Obtain the samples pictures for training preset model；The preset model is the primary dcreening operation network mould of original image for identification
Type；
Classify to all samples pictures, and according to classification host type construct respectively include similar samples pictures training
Collection；Wherein, it is described classification host type include interfere picture, shooting exterior surface do not include off-note first object picture,
The second Target Photo that exterior surface includes the off-note is shot, the off-note includes protruding features and/or specifies
Color characteristic.
2. the method according to claim 1, wherein the corresponding classification host type of the interference picture further includes one
The overseas tag along sort of class and the overseas tag along sort of two classes；Correspondingly, the method also includes:
According to the overseas tag along sort of described one kind and the overseas tag along sort of two classes, construct respectively corresponding with the interference picture
The sub- training set of training set；Wherein, the overseas tag along sort of one kind be shooting defect based on the original image, with it is to be checked
Survey what the unrelated shooting position of target site determined；The overseas tag along sort of two classes be worth based on no medical judgment it is original
Picture, the original image for being attached with covering, include slaking residue object original image determine.
3. method according to claim 1 or 2, which is characterized in that the corresponding classification host type of the first object picture
It further include first object picture classification label, hole shape structure first object picture classification label, base based on partial structurtes feature
In the first object picture classification label of global structure feature；Correspondingly, the method also includes:
According to based on partial structurtes feature first object picture classification label, hole shape structure first object picture classification label,
First object picture classification label based on global structure feature constructs training set corresponding with the first object picture respectively
Sub- training set；Wherein, the first object picture classification label based on partial structurtes feature be based on partially with shape and/
Or the first object picture of tone variations is the first object picture classification label based on global structure feature determine, described
It is determined based on the global first object picture with stomach corner structure and/or texture structure.
4. according to the method described in claim 2, it is characterized in that, the covering includes:
Lumps suspended matter, bubble population and mucous membrane body.
5. according to the method described in claim 2, it is characterized in that, the target site to be detected is stomach.
6. method according to claim 1 or 2, which is characterized in that the primary dcreening operation network model is convolutional neural networks
inceptionV3。
7. a kind of construction device of training set characterized by comprising
Acquiring unit, for obtaining the samples pictures for training preset model；The preset model is original graph for identification
The primary dcreening operation network model of piece；
Construction unit, for classifying to all samples pictures, and being constructed respectively according to classification host type includes similar sample
The training set of this picture；Wherein, the classification host type includes that picture, shooting exterior surface is interfered not to include the of off-note
One Target Photo, shooting exterior surface include the second Target Photo of the off-note, and the off-note includes that protrusion is special
Sign and/or designated color feature.
8. a kind of electronic equipment characterized by comprising processor, memory and bus, wherein
The processor and the memory complete mutual communication by the bus；
The memory is stored with the program instruction that can be executed by the processor, and the processor calls described program to instruct energy
Enough methods executed as described in claim 1 to 6 is any.
9. a kind of non-transient computer readable storage medium, which is characterized in that the non-transient computer readable storage medium is deposited
Computer instruction is stored up, the computer instruction makes the computer execute the method as described in claim 1 to 6 is any.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910251169.7A CN110097081A (en) | 2019-03-29 | 2019-03-29 | A kind of construction method and device of training set |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910251169.7A CN110097081A (en) | 2019-03-29 | 2019-03-29 | A kind of construction method and device of training set |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110097081A true CN110097081A (en) | 2019-08-06 |
Family
ID=67444186
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910251169.7A Pending CN110097081A (en) | 2019-03-29 | 2019-03-29 | A kind of construction method and device of training set |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110097081A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106709511A (en) * | 2016-12-08 | 2017-05-24 | 华中师范大学 | Urban rail transit panoramic monitoring video fault detection method based on depth learning |
| CN106910184A (en) * | 2017-01-12 | 2017-06-30 | 杭州电子科技大学 | Endoscopic images intestinal bleeding detection method based on depth convolutional neural networks |
| CN108038850A (en) * | 2017-12-08 | 2018-05-15 | 天津大学 | A kind of drainage pipeline Exception Type automatic testing method based on deep learning |
-
2019
- 2019-03-29 CN CN201910251169.7A patent/CN110097081A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106709511A (en) * | 2016-12-08 | 2017-05-24 | 华中师范大学 | Urban rail transit panoramic monitoring video fault detection method based on depth learning |
| CN106910184A (en) * | 2017-01-12 | 2017-06-30 | 杭州电子科技大学 | Endoscopic images intestinal bleeding detection method based on depth convolutional neural networks |
| CN108038850A (en) * | 2017-12-08 | 2018-05-15 | 天津大学 | A kind of drainage pipeline Exception Type automatic testing method based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| 郑树泉等: "《工业智能技术与应用》", 31 January 2019, 上海科学技术出版社 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107392091B (en) | Agricultural artificial intelligence crop detection method, mobile terminal and computer readable medium | |
| CN109241967B (en) | Thyroid ultrasound image automatic identification system based on deep neural network, computer equipment and storage medium | |
| CN110232383A (en) | A kind of lesion image recognition methods and lesion image identifying system based on deep learning model | |
| CN108596046A (en) | A kind of cell detection method of counting and system based on deep learning | |
| CN109063740A (en) | The detection model of ultrasonic image common-denominator target constructs and detection method, device | |
| US11819347B2 (en) | Dental imaging system utilizing artificial intelligence | |
| CN108615046A (en) | A kind of stored-grain pests detection recognition methods and device | |
| CN108899075A (en) | A kind of DSA image detecting method, device and equipment based on deep learning | |
| CN110084275A (en) | A kind of choosing method and device of training sample | |
| CN110097080A (en) | A kind of construction method and device of tag along sort | |
| CN113052843B (en) | Method, apparatus, system, storage medium and computing device for assisting endoscopy | |
| CN110110750A (en) | A kind of classification method and device of original image | |
| CN110110749A (en) | Image processing method and device in a kind of training set | |
| CN110097081A (en) | A kind of construction method and device of training set | |
| CN110070112A (en) | A kind of method and device of determining tag along sort | |
| CN110083727A (en) | A kind of method and device of determining tag along sort | |
| CN110097082A (en) | A kind of method for splitting and device of training set | |
| CN110096604A (en) | A kind of method and device of determining tag along sort | |
| CN114581402A (en) | Capsule endoscope quality inspection method, device and storage medium | |
| CN110084277B (en) | Splitting method and device of training set | |
| CN110110747A (en) | A kind of method and device of determining tag along sort | |
| CN110070113B (en) | Training method and device for training set | |
| CN112634266A (en) | Semi-automatic marking method, medium, equipment and device for laryngoscope image | |
| CN110084276B (en) | Splitting method and device of training set | |
| CN110084280A (en) | A kind of method and device of determining tag along sort |
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: 20190806 |