CN108447049A - A kind of digitlization physiology organism dividing method fighting network based on production - Google Patents
A kind of digitlization physiology organism dividing method fighting network based on production Download PDFInfo
- Publication number
- CN108447049A CN108447049A CN201810163158.9A CN201810163158A CN108447049A CN 108447049 A CN108447049 A CN 108447049A CN 201810163158 A CN201810163158 A CN 201810163158A CN 108447049 A CN108447049 A CN 108447049A
- Authority
- CN
- China
- Prior art keywords
- organism
- medical image
- data
- training
- active contour
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/149—Segmentation; Edge detection involving deformable models, e.g. active contour models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/20112—Image segmentation details
- G06T2207/20116—Active contour; Active surface; Snakes
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention discloses a kind of digitlization physiology organism dividing methods fighting network based on production, include the following steps:Format conversion is carried out to the original medical image of organism first;Secondly active contour method is used, it is semi-automatic that area-of-interest is split to part biological body medical image, it is used as the data of production confrontation network training;The organism medical image that original organism medical image and active contour method are divided is sent into confrontation type generation network again and is trained study, obtains a training pattern;The model finally obtained using training study, generates the corresponding segmentation image of new organism medical image.The invention avoids traditional physiologic images dividing methods the problems such as locality, boundary is interrupted, autgmentability is poor easily occurs, and the complete effective expression of healthy organism physiological structure is made to be equalized on the critical index such as segmentation precision, efficiency, stability, robustness.
Description
Technical field
The present invention relates to image procossing medical domains, more particularly to a kind of digital fighting network based on production
Manage organism dividing method.
Background technology
With the development of medical imaging and computer aided technique, image Segmentation Technology has become Medical Image Processing
Research hotspot.Researcher is dedicated to new Medical image segmentation algorithm research on one side, and continuously attempting to again on one side will be different
Method is cleverly combined, and is more effectively divided to medical image with reaching.
Some Automatic medical image segmentation technologies are come into being, and such as threshold method, region-growing method, edge detection method, are based on
Method, fuzzy clustering algorithm, the method based on genetic algorithm, the method etc. based on wavelet transformation of movable contour model, these sides
Method is the representative of medical image segmentation field achievement.Threshold method is a kind of simple and effective method, especially for background with
The larger image of target area contrast, segmentation result is even more ideal, and the threshold value in algorithm needs in cutting procedure constantly
It manually adjusts and improves, so being interactive mostly, judged on the basis that user's vision is estimated;It gives birth in region
Regular way requires first selected seed pixel, and then by the region where potting gum similar with its to it, basic principle is just
It is that similar pixel is assembled into region, to noise-sensitive, if the selection of sub-pixel point is improper, segmentation result just will appear mistake
Accidentally, and close for gray value in image but non-conterminous multiple regions once cannot all be split;Edge detection method
It is split by the detection to target area boundaries, the discontinuous part of gray value in image is first found, by unevenness
The segmentation to image is realized in the detection of even interregional intersection;It, will based on the method for movable contour model by energy minimization
One initial curve with energy function is gradually deformed and is moved towards objective contour direction to be detected, finally converges to mesh
Boundary is marked, a smooth and continuous profile is obtained;Fuzzy set theory is combined by fuzzy clustering algorithm with clustering algorithm, is obscured
Collection theory has preferable descriptive power to the uncertainty of image, is easy to apply, but the data of wherein each pixel are mutually solely
Vertical, do not utilize the spatial information of image;Genetic algorithm is since an initializaing variable group, by the gene in chromosome
It is operable to complete by for optimizing, until algorithmic statement finds optimal segmenting threshold, is finally completed the segmentation of different zones;It is small
Wave conversion is to the Fourier successions analyzed and development, and the basic skills using wavelet transformation progress medical image segmentation is logical
The coefficient that image histogram is resolved into different stage by wavelet transformation is crossed, is controlled with scale and according to wavelet coefficient and given point
Criterion is cut to select threshold value.
The image segmentation of organism physiological structure be the key that realize digital virtual physiology biological standard map construction because
Element and technological difficulties.Organism multi-source physiologic images data imaging principle is different, and image degradation model is totally different.CT image volumes are imitated
Organization edge part should be caused there are ambiguity, accuracy reduces;Ultrasonoscopy signal-to-noise ratio is relatively low;It can not in MRI image
Great challenge is brought the gray-scale deviation field avoided to image segmentation.
Easily there is the problems such as locality, boundary is interrupted, autgmentability is poor in traditional physiologic images dividing method.Meanwhile biology
Body sample respectively organizes organ structure there are individual difference, and the complete effective expression needs of healthy organism physiological structure are being divided
Give equilibrium on the critical index such as precision, efficiency, stability, robustness.
Invention content
The purpose of the present invention is to propose to a kind of digitlization physiology organism dividing methods fighting network based on production, will
Production confrontation network is introduced into the segmentation of organism digitlization physiologic images, learns original life using generator and arbiter
The relationship of reason image and its segmentation image simultaneously generates model, and the segmentation knot of new physiologic images is generated using the model of generation
Fruit, to make up the deficiencies in the prior art.
In order to achieve the above objectives, the present invention is achieved through the following technical solutions:
A kind of digitlization physiology organism dividing method being fought network based on production, is included the following steps:
(1) organism medical image sequences I is obtainedi, and intercept the valid interval of pixel value;
(2) active contour method is used, it is semi-automatic that organism medical image is split, it is used as production confrontation network
Trained data Ii′;
(3) by original organism medical image IiThe organism medical image I divided with active contour methodi' be sent into
Confrontation type generates network and is trained study, obtains a training pattern;
(4) training pattern for utilizing step (3) to obtain, generates the corresponding segmentation image of new organism medical image.
Further, in step (2), the active contour method includes the following steps:It is put near interesting target
Set an initial profile line C;Calculate the vertical normal vector of the contour lineThe internal energy of profile is acted in normal direction
The summation F of (internal force) and external energy (external force);According to partial differential equationIn internal energy (internal force) and outside
External energy is deformed under the action of energy (external force) attracts active contour to be moved towards object edge, and internal energy keeps castor
Wide slickness and topological, when energy reaches minimum, active contour converges to object edge to be detected.
Further, in step (3), the confrontation type generates network and includes the following steps:By active contour method point
The organism medical image I cuti' it is used as constraint y, y and noise z are sent into generator G together generates data, generates
Device G will minimize the loss of D;Organism medical image IiAs input while it being sent into arbiter D with constraint y, is generated
Cross-domain vector, and further judge organism medical image IiIt is the probability of true training data, training pattern maximum probability
Divide to authentic specimen;G and D training simultaneously, wants fixed party in training, updates the parameter of another party, alternating iteration makes other side's
Mistake maximizes;The accuracy rate of final discrimination model is equal to 50%, and entire model state reaches Nash Equilibrium;Finally obtain training
Good model.
Further, in G and D simultaneously training, fixed party in training updates the parameter of another party, alternating iteration makes
The mistake of other side maximizes;Optimization process is " game of a binary minimax " problem, and object function is as follows:
Wherein, Pdata(Ii, y) and indicate that the truthful data in view of constraint y is distributed;Pdata(Ii) indicate not consider item
Part constrains the truthful data distribution of y;pz(Z) prior distribution is indicated;E () indicates to calculate desired value;D(Ii, y) and indicate IiSource
In truthful data rather than the probability of generation data, when input data is sampled from truthful data IiWhen, the target of D is so that output is general
Rate value D (Ii,Ii') level off to 1;When input carrys out self-generating data G (Ii', z) when, the target of D is correctly to judge data source, is made
Obtain D (Ii,G(Ii', z)) level off to 0, while the target of G is so that it levels off to 1.
Advantages of the present invention and technique effect:
The present invention is labeled segmentation using the semi-automatic organism medical image to as training set of active contour method,
Semi-automatic method improves the accuracy and timeliness for doing mark segmentation;Automatic mesh generation organism is generated using confrontation type to cure
The segmentation result for learning image, avoids traditional physiologic images dividing method and locality, boundary is interrupted, autgmentability is poor etc. easily occurs
Problem keeps the complete effective expression of healthy organism physiological structure key in segmentation precision, efficiency, stability, robustness etc.
It is equalized in index.
Organism medical image cutting method provided by the invention realizes each group of organism with information technology in real time
The segmentation knitted improves the speed and precision of segmentation, further to push nutritional need analysis, medicament residue detection and efficiently matching
The development for closing bait provides some theoretical reference foundations, for find influence of the environmental stress to organism physiological function provide effectively according to
According to.
Description of the drawings
Fig. 1 is the overall flow figure of the present invention.
Fig. 2 is two-dimentional fish body MRI image in specific embodiment.
Fig. 3 is two-dimentional fish body MRI images of the Fig. 2 after the segmentation of active contour method.
Fig. 4 is that Fig. 2 generates the image that network generates under the conditions of Fig. 3 by confrontation type.
Fig. 5 is newly-increased two-dimentional fish body MRI image.
Fig. 6 is two-dimentional fish body MRI images of the Fig. 5 after the segmentation of active contour method.
Fig. 7 is the segmentation image that Fig. 5 generates that network is generated by confrontation type.
Specific implementation mode
To make present disclosure and advantage be more clear, below by way of specific embodiment, and it is described with reference to the accompanying drawings
The specific implementation process of the present invention.
Mariculture industry is one of the pillar industry of China's blue economic development.Flounder flounder class factoryization cultivation in China is from starting
So far only 20 years have suddenly become the first in the world flounder flounder class cultivation big country.According to national flounder flounder class industrial system investigation statistics
(2009), ten thousand tons of China flounder flounder class gross annual output amount 7.9-8.9, wherein Bohai Rim main producing region account for the 92% of national total output, production
It measures maximum 3 kinds and is followed successively by turbot, lefteye flounder and Cynoglossus semilaevis, very important ground is occupied in marine fish culture industry
Position.With the growth of cultivation scale and intensity, healthy aquaculture problem is concerned." more treasured fish disturbance " in 2006 is due to portion
Divide raiser to abuse antibiotic and causes the medicament residue problems such as chloramphenicol, Ciprofloxacin, malachite chlorine, adult fish color and luster, the bodily form, fertilizer
Full scale, colloid thickness decline at rate, vigor and mouthfeel quality, and drug resistance enhancing, immunity reduce, and are once leading to the areas flounder Die Leizhuyang
Aquatic products are unsalable, and negative effect and crushing blow are caused to entire industry.Therefore commenting for flounder flounder class health fish body resource
It is vital to estimate.
The present embodiment is described in detail with the MRI image segmentation of this bastard halibut and plaice of turbot.
The overall flow of the present invention is as shown in Figure 1, detailed process is as follows:
(1) MRI image format is converted
Turbot fish body MRI image sequence image is obtained from ct apparatus or database, intercepts picture
The valid interval of element value, is converted into common Computer Image Processing format;Two-dimentional turbot fish body MRI figures after format conversion
As I is as shown in Figure 2.
(2) active contour method semi-automatic segmentation is used to make data set
It is semi-automatic that part biological body medical image is split using active contour method, it is used as production confrontation net
The data of network training;By fish body medical image I in sequence IiPixel be divided into bone Ii 1, fish intestines Ii 2, fatty Ii 3, fish body other
Position Ii 4With the other parts I in addition to fish bodyi 5;Include the following steps:
A) it is placed around an initial profile line C in interesting target;
B) the vertical normal vector of the contour line is calculated
C) the summation F of the internal energy (internal force) and external energy (external force) of profile is acted in normal direction
Wherein gIIt is from fish body medical image IiGradient amplitude derived from velocity function, k is the average curvature of profile, α,
β, γ are the weights of the relative contribution for three components for modulating F;
D) according to partial differential equationBecome under the action of internal energy (internal force) and external energy (external force)
Shape external energy attracts active contour to be moved towards object edge, and internal energy keeps the slickness and topological of active contour,
When energy reaches minimum, active contour converges to the turbot two dimension fish body after object edge segmentation mark to be detected
MRI image Ii', as shown in Figure 3.
(3) production confrontation network training is used to divide turbot two dimension fish body MRI image
By original fish body medical image IiThe fish body medical image I divided with active contour methodi' it is sent into confrontation type
It generates network and is trained study, obtain a training pattern;Include the following steps:
A) the fish body medical image I for dividing active contour methodi' it is used as constraint y, together by y and noise z
It is sent into generator G and generates data, generator G will minimize the loss of D;
B) fish body medical image IiAs input while it being sent into arbiter D with constraint y, generates cross-domain vector, and
Further judge fish body medical image IiIt is the probability of true training data, divides to authentic specimen to training pattern maximum probability;
C) G and D training simultaneously, but fixed party in training, update the parameter of another party, alternating iteration makes other side's
Mistake maximizes;Optimization process is " game of a binary minimax " problem, and object function is as follows:
Wherein, Pdata(Ii, y) and indicate that the truthful data in view of constraint y is distributed;Pdata(Ii) indicate not consider item
Part constrains the truthful data distribution of y;pz(Z) prior distribution is indicated;E () indicates to calculate desired value;D(Ii, y) and indicate IiSource
In truthful data rather than the probability of generation data, when input data is sampled from truthful data IiWhen, the target of D is so that output is general
Rate value D (Ii,Ii') level off to 1;When input carrys out self-generating data G (Ii', z) when, the target of D is correctly to judge data source, is made
Obtain D (Ii,G(Ii', z)) level off to 0, while the target of G is so that it levels off to 1.
The optimization problem that confrontation type generates network is a minimum-maximization problem, and confrontation type generates the target letter of network
Number can be described as follows:
The method that we use alternative optimization:First fix generator G, optimization arbiter D so that the differentiation accuracy rate of D is most
Bigization;Then arbiter D, optimization generator G so that the differentiation accuracy rate of D minimizes are fixed;
E) accuracy rate of final discrimination model is equal to 50%, and entire model state reaches Nash Equilibrium;
F) trained model is finally obtained;
The turbot two dimension fish body MRI image divided using production confrontation network training is as shown in Figure 4.
(4) model that step (3) generates is utilized to generate the segmentation result of newly-increased turbot two dimension fish body MRI image.
Input increases turbot two dimension fish body MRI image newly, as shown in figure 5, using the model parameter generated, generates new point
It cuts as a result, as shown in Figure 7.It is compared using the image 6 that active contour method semi-automatic segmentation obtains with us, it can be seen that we
The segmentation result effect that network obtains is generated using confrontation type to become apparent.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
The change or replacement expected without creative work, should be covered by the protection scope of the present invention.Therefore, of the invention
Protection domain should be determined by the scope of protection defined in the claims.
Claims (4)
1. a kind of digitlization physiology organism dividing method fighting network based on production, which is characterized in that the dividing method
Include the following steps:
(1) organism medical image sequences I is obtainedi, and intercept the valid interval of pixel value;
(2) active contour method is used, it is semi-automatic that organism medical image is split, it is used as production confrontation network training
Data Ii′;
(3) by original organism medical image IiThe organism medical image I divided with active contour methodi' it is sent into confrontation
Formula generates network and is trained study, obtains a training pattern;
(4) training pattern for utilizing step (3) to obtain, generates the corresponding segmentation image of new organism medical image.
2. dividing method as described in claim 1, which is characterized in that in the step (2), the active contour method includes
Following steps:It is placed around an initial profile line C in interesting target;Calculate the vertical normal vector of the contour lineIn method
Line direction acts on the internal energy of profile and the summation F of external energy;According to partial differential equationIn internal energy
Deformation external energy attracts active contour to be moved towards object edge under the action of amount and external energy, and internal energy holding activity
The slickness and topological of profile, when energy reaches minimum, active contour converges to object edge to be detected.
3. dividing method as described in claim 1, which is characterized in that in the step (3), the confrontation type generates network packet
Include following steps:The organism medical image I that active contour method is dividedi' it is used as constraint y, by y and noise z
It is sent into generator G together and generates data, generator G will minimize the loss of D;Organism medical image IiWith constraint y
As input while it being sent into arbiter D, generates cross-domain vector, and further judge organism medical image IiIt is true training number
According to probability, training pattern maximum probability point to authentic specimen;G and D training simultaneously, wants fixed party, update another in training
The parameter of one side, alternating iteration make the mistake of other side maximize;The accuracy rate of final discrimination model is equal to 50%, entire model
State reaches Nash Equilibrium;Finally obtain trained model.
4. dividing method as claimed in claim 3, which is characterized in that in the G and D simultaneously training, one is fixed in training
Side, updates the parameter of another party, and alternating iteration makes the mistake of other side maximize;Optimization process is one, and " binary minimax is rich
Play chess " problem, object function is as follows:
Wherein, Pdata(Ii, y) and indicate that the truthful data in view of constraint y is distributed;Pdata(Ii) indicate not consider constraint
The truthful data of y is distributed;pz(Z) prior distribution is indicated;E () indicates to calculate desired value;D(Ii, y) and indicate IiFrom true
Data rather than the probability for generating data, when input data is sampled from truthful data IiWhen, the target of D is so that output probability value D
(Ii,Ii') level off to 1;When input carrys out self-generating data G (Ii', z) when, the target of D is correctly to judge data source so that D
(Ii,G(Ii', z)) level off to 0, while the target of G is so that it levels off to 1.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810163158.9A CN108447049A (en) | 2018-02-27 | 2018-02-27 | A kind of digitlization physiology organism dividing method fighting network based on production |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810163158.9A CN108447049A (en) | 2018-02-27 | 2018-02-27 | A kind of digitlization physiology organism dividing method fighting network based on production |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108447049A true CN108447049A (en) | 2018-08-24 |
Family
ID=63192678
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810163158.9A Pending CN108447049A (en) | 2018-02-27 | 2018-02-27 | A kind of digitlization physiology organism dividing method fighting network based on production |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108447049A (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109523014A (en) * | 2018-10-22 | 2019-03-26 | 广州大学 | News comment automatic generation method and system based on production confrontation network model |
CN109656878A (en) * | 2018-12-12 | 2019-04-19 | 中电健康云科技有限公司 | Health account data creation method and device |
CN109671084A (en) * | 2018-11-15 | 2019-04-23 | 华东交通大学 | A kind of measurement method of workpiece shapes |
CN109857845A (en) * | 2019-01-03 | 2019-06-07 | 北京奇艺世纪科技有限公司 | Model training and data retrieval method, device, terminal and computer readable storage medium |
CN110074813A (en) * | 2019-04-26 | 2019-08-02 | 深圳大学 | A kind of ultrasonic image reconstruction method and system |
CN111568376A (en) * | 2020-05-11 | 2020-08-25 | 四川大学 | Direct three-dimensional scanning method and system for physiological motion boundary of oral soft tissue |
CN111652887A (en) * | 2020-05-13 | 2020-09-11 | 腾讯科技(深圳)有限公司 | Image segmentation model training method and device, computer equipment and storage medium |
CN111815656A (en) * | 2020-07-22 | 2020-10-23 | 北京字节跳动网络技术有限公司 | Video processing method, video processing device, electronic equipment and computer readable medium |
CN111985394A (en) * | 2020-08-19 | 2020-11-24 | 东南大学 | Semi-automatic instance labeling method and system for KITTI data set |
CN111985510A (en) * | 2019-05-21 | 2020-11-24 | 北京新唐思创教育科技有限公司 | Generative model training method, image generation device, medium, and terminal |
CN113569855A (en) * | 2021-07-07 | 2021-10-29 | 江汉大学 | Tongue picture segmentation method, equipment and storage medium |
CN113674330A (en) * | 2021-07-12 | 2021-11-19 | 华南理工大学 | Pseudo CT image generation system based on generation countermeasure network |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1584936A (en) * | 2004-06-07 | 2005-02-23 | 东软飞利浦医疗设备系统有限责任公司 | Three-dimensional dividing method for medical images |
CN103646398A (en) * | 2013-12-04 | 2014-03-19 | 山西大学 | Demoscopy focus automatic segmentation method |
CN107220980A (en) * | 2017-05-25 | 2017-09-29 | 重庆理工大学 | A kind of MRI image brain tumor automatic division method based on full convolutional network |
-
2018
- 2018-02-27 CN CN201810163158.9A patent/CN108447049A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1584936A (en) * | 2004-06-07 | 2005-02-23 | 东软飞利浦医疗设备系统有限责任公司 | Three-dimensional dividing method for medical images |
CN103646398A (en) * | 2013-12-04 | 2014-03-19 | 山西大学 | Demoscopy focus automatic segmentation method |
CN107220980A (en) * | 2017-05-25 | 2017-09-29 | 重庆理工大学 | A kind of MRI image brain tumor automatic division method based on full convolutional network |
Non-Patent Citations (3)
Title |
---|
JAEMIN SON ET AL.: "Retinal Vessel Segmentation in Fundoscopic Images with Generative Adversarial Networks", 《ARXIV》 * |
MEHDI MIRZA ET AL.: "Conditional Generative Adversarial Nets", 《ARXIV》 * |
基于活动轮廓模型的医学图像分割关键技术研究: "基于活动轮廓模型的医学图像分割关键技术研究", 《中国博士学位论文全文数据库信息科技辑》 * |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109523014A (en) * | 2018-10-22 | 2019-03-26 | 广州大学 | News comment automatic generation method and system based on production confrontation network model |
CN109671084A (en) * | 2018-11-15 | 2019-04-23 | 华东交通大学 | A kind of measurement method of workpiece shapes |
CN109671084B (en) * | 2018-11-15 | 2023-05-30 | 华东交通大学 | Method for measuring shape of workpiece |
CN109656878A (en) * | 2018-12-12 | 2019-04-19 | 中电健康云科技有限公司 | Health account data creation method and device |
CN109857845A (en) * | 2019-01-03 | 2019-06-07 | 北京奇艺世纪科技有限公司 | Model training and data retrieval method, device, terminal and computer readable storage medium |
CN110074813B (en) * | 2019-04-26 | 2022-03-04 | 深圳大学 | Ultrasonic image reconstruction method and system |
CN110074813A (en) * | 2019-04-26 | 2019-08-02 | 深圳大学 | A kind of ultrasonic image reconstruction method and system |
CN111985510A (en) * | 2019-05-21 | 2020-11-24 | 北京新唐思创教育科技有限公司 | Generative model training method, image generation device, medium, and terminal |
CN111985510B (en) * | 2019-05-21 | 2022-05-31 | 北京新唐思创教育科技有限公司 | Generative model training method, image generation device, medium, and terminal |
CN111568376A (en) * | 2020-05-11 | 2020-08-25 | 四川大学 | Direct three-dimensional scanning method and system for physiological motion boundary of oral soft tissue |
CN111652887A (en) * | 2020-05-13 | 2020-09-11 | 腾讯科技(深圳)有限公司 | Image segmentation model training method and device, computer equipment and storage medium |
CN111652887B (en) * | 2020-05-13 | 2023-04-07 | 腾讯科技(深圳)有限公司 | Image segmentation model training method and device, computer equipment and storage medium |
CN111815656A (en) * | 2020-07-22 | 2020-10-23 | 北京字节跳动网络技术有限公司 | Video processing method, video processing device, electronic equipment and computer readable medium |
CN111815656B (en) * | 2020-07-22 | 2023-08-11 | 抖音视界有限公司 | Video processing method, apparatus, electronic device and computer readable medium |
CN111985394A (en) * | 2020-08-19 | 2020-11-24 | 东南大学 | Semi-automatic instance labeling method and system for KITTI data set |
CN113569855A (en) * | 2021-07-07 | 2021-10-29 | 江汉大学 | Tongue picture segmentation method, equipment and storage medium |
CN113674330A (en) * | 2021-07-12 | 2021-11-19 | 华南理工大学 | Pseudo CT image generation system based on generation countermeasure network |
CN113674330B (en) * | 2021-07-12 | 2023-02-14 | 华南理工大学 | Pseudo CT image generation system based on generation countermeasure network |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108447049A (en) | A kind of digitlization physiology organism dividing method fighting network based on production | |
CN105469100B (en) | Skin biopsy image pathological characteristics recognition methods based on deep learning | |
CN106778902A (en) | Milk cow individual discrimination method based on depth convolutional neural networks | |
CN105957066B (en) | CT image liver segmentation method and system based on automatic context model | |
Klingenberg | Analyzing fluctuating asymmetry with geometric morphometrics: concepts, methods, and applications | |
CN107977671A (en) | A kind of tongue picture sorting technique based on multitask convolutional neural networks | |
CN107358626A (en) | A kind of method that confrontation network calculations parallax is generated using condition | |
CN106682616A (en) | Newborn-painful-expression recognition method based on dual-channel-characteristic deep learning | |
CN106446942A (en) | Crop disease identification method based on incremental learning | |
CN108764286A (en) | The classifying identification method of characteristic point in a kind of blood-vessel image based on transfer learning | |
Du et al. | Image-based high-throughput detection and phenotype evaluation method for multiple lettuce varieties | |
CN106530283A (en) | SVM (support vector machine)-based medical image blood vessel recognition method | |
CN108765430A (en) | A kind of heart left chamber region segmentation method based on cardiac CT image and machine learning | |
CN107292312A (en) | Tumour recognition methods | |
CN116563707B (en) | Lycium chinense insect pest identification method based on image-text multi-mode feature fusion | |
CN112949517B (en) | Plant stomata density and opening degree identification method and system based on deep migration learning | |
Hanschen et al. | Evolution of individuality: a case study in the volvocine green algae | |
CN114119872A (en) | Method for analyzing 3D printing intraspinal plants based on artificial intelligence big data | |
Rappaport et al. | The emergence of religion in human evolution | |
CN106296702A (en) | Cotton Images dividing method and device under natural environment | |
CN109919216B (en) | Counterlearning method for computer-aided diagnosis of prostate cancer | |
CN108520525A (en) | A kind of spinal cord dividing method based on convex constraint seed region growth | |
CN111986216B (en) | RSG liver CT image interactive segmentation algorithm based on neural network improvement | |
CN106909881A (en) | The method and system of corn breeding base ridge number are extracted based on unmanned aerial vehicle remote sensing images | |
CN117618723A (en) | Be used for improving children's attention training system |
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 |
Application publication date: 20180824 |
|
RJ01 | Rejection of invention patent application after publication |