CN108596840B - Data set enhancement method for deep learning evaluation of vascular network development level - Google Patents
Data set enhancement method for deep learning evaluation of vascular network development level Download PDFInfo
- Publication number
- CN108596840B CN108596840B CN201810261472.0A CN201810261472A CN108596840B CN 108596840 B CN108596840 B CN 108596840B CN 201810261472 A CN201810261472 A CN 201810261472A CN 108596840 B CN108596840 B CN 108596840B
- Authority
- CN
- China
- Prior art keywords
- picture
- image
- point
- data set
- data
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 27
- 230000002792 vascular Effects 0.000 title claims abstract description 21
- 238000013135 deep learning Methods 0.000 title claims abstract description 20
- 238000011161 development Methods 0.000 title abstract description 9
- 238000011156 evaluation Methods 0.000 title description 3
- 230000003321 amplification Effects 0.000 claims abstract description 31
- 238000003199 nucleic acid amplification method Methods 0.000 claims abstract description 31
- 238000007781 pre-processing Methods 0.000 claims abstract description 11
- 238000012549 training Methods 0.000 claims description 17
- 230000001965 increasing effect Effects 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 7
- 238000005192 partition Methods 0.000 claims description 6
- 239000011521 glass Substances 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims 2
- 230000002708 enhancing effect Effects 0.000 abstract description 2
- 210000004204 blood vessel Anatomy 0.000 description 15
- 230000018109 developmental process Effects 0.000 description 7
- 210000001161 mammalian embryo Anatomy 0.000 description 4
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000013020 embryo development Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- 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
- G06T2207/30004—Biomedical image processing
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
A data set enhancement method for deep learning assessment of vascular network developmental levels, comprising the steps of: a, preprocessing an original image; b, performing data amplification on the preprocessed image; c, dividing N rectangular regions with equal areas in the space range of the original image; and d, randomly selecting the same proportion of pictures in each area to generate a data set. The invention provides a data set enhancing method for deep learning assessment of the development level of a vascular network, which has more comprehensive and balanced data set distribution and enhances the robustness of a model.
Description
Technical Field
The invention relates to a data set enhancement method, which can provide an image data set for classifying a blood vessel network by applying deep learning and belongs to the technical field of deep learning.
Background
Assessment of embryonic development stage and development is of great significance for physiological and clinical studies. The local vascular network morphology and topological pattern can be used for evaluating the growth and development conditions of the embryo. The deep learning method can evaluate the development level of the embryo by learning the local vascular network morphology and topological patterns of various development stages, and has better performance compared with the traditional method.
Because the embryonic vascular network structure has spatial heterogeneity, that is, the difference of the vascular network morphological structures at different spatial positions is large, the local vascular network images at different spatial positions need to be intercepted and input into the deep learning model, so that the performance of the model is ensured. The method of line-scanning and intercepting partial images in a complete blood vessel network image cannot obtain enough image data. The method for randomly intercepting the partial images in the complete blood vessel network image can ensure that a data set has a large enough scale and is suitable for deep learning, but is difficult to ensure that the selected images uniformly cover a complete embryo blood vessel network area, and the similarity between the partially selected images is too high, so that the selected data set cannot completely represent the structural characteristics of the embryo blood vessel network, and the performance of a model is influenced.
The data set enhancement method for deep learning assessment of the development level of the vascular network can solve the problems of insufficient data volume of vascular network pictures, excessive similarity among the pictures after image data amplification and uneven distribution of data set selection picture space. The method is used for deep learning data set enhancement of the classification of the blood vessel network.
Disclosure of Invention
In order to overcome the defects of insufficient image data set quantity, overhigh image similarity after data amplification and uneven data set selected image space distribution in the prior art, the invention provides a data set enhancement method for deep learning evaluation of the vascular network development level, which has more comprehensive and balanced data set distribution and enhanced model robustness.
In order to solve the technical problems, the invention adopts the following technical scheme:
a data set enhancement method for deep learning assessment of vascular network developmental levels, comprising the steps of:
a. preprocessing an original image;
b. performing data amplification on the preprocessed image;
c. dividing N rectangular regions with equal area in the space range of an original image;
d. and randomly selecting the same proportion of pictures in each area to generate a data set.
Further, in the step a, the pretreatment comprises the following steps:
a1 removing the area of invalid information around the image;
a2, zooming the images generated by different magnifying glass multiples to reach the same scale standard;
a3 binarizes the image to remove noise point interference.
Invalid regions and noise points in the original blood vessel network image can be effectively reduced through the steps.
Still further, in the step b, the data amplification comprises the following steps:
b1 setting a picture coverage threshold E, wherein the threshold is set to avoid the overlarge overlapping area of the amplified pictures;
b2 uniform random spotting is performed on the image after pre-processing. And (3) generating a picture by taking the uniformly and randomly acquired coordinate point as a central point of the amplification picture, calculating the coverage rate of the picture and each existing picture, assuming that the coordinate of the uniformly and randomly acquired point is (X, Y), generating a square picture by taking the point as the central point of the picture, wherein the side length of the picture is C, the coordinate of the other point is (X ', Y'), the side length of the picture is C, and the overlapping area of the two pictures is as follows:
the coverage of these two graphs was calculated as:
b3, if the coverage rate calculated by the picture generated by uniformly and randomly taking the point and the existing picture is lower than the coverage rate threshold value E, the picture generated by the point as the picture center point meets the requirement and is stored, otherwise, the random and uniform point taking is carried out again, and the operation is repeated until the preprocessed picture can not take the coordinate point meeting the condition;
b4 rotates and mirrors the image saved in b3 by a certain angle to generate new data, generates a new data every time rotating by an angle, and generates a new data every time mirroring is performed similarly.
Through the steps, the data volume can be effectively increased, and the generated new data can avoid overlarge overlapped areas among the pictures, so that the over-fitting condition is avoided.
Further, in the step c, the area division includes the following steps:
the division of the c1 area depends on the size of the picture after pretreatment and the size of the picture after amplification, and the picture after pretreatment is divided into N rectangular areas with equal area;
c2, if the size of the preprocessed picture is far larger than that of the data amplified picture, properly increasing the divided rectangular areas; otherwise, the divided rectangular area is reduced appropriately.
Preferably, in the step c2, the rectangular regions are divided in various ways, which may be parallel to the length or width of the picture, or may be divided into grids, but each rectangular region with the same area is ensured to have enough expansion of the center point of the picture, so as to facilitate the selection of the picture.
The reference of the selection area is provided for generating a uniform data set for the subsequent partition selection through the steps.
In the step d, generating the data set includes the following steps:
d1 setting a ratio, G%, which is used to determine the number of amplifications data for each partition into the training set;
d2 step c the preprocessed image has been divided into N rectangular regions of equal area, denoted A1,A2……ANShowing the divided areas, if the central coordinate point of F pictures falls on A1Region, then we are at A1Randomly selecting [ F X G%]([]Representing to get a whole) of the amplified pictures generated by the central coordinate point to a training set, and putting the rest pictures into a testing set; to A2…ANThe region repeats the above operation.
Through the steps, the generated data set can be distributed more uniformly, and more characteristic information can be learned by the model.
The invention has the following beneficial effects: data amplification is carried out under the condition of limited data quantity, and the data set after amplification is distributed more comprehensively and uniformly, so that the robustness of the model is enhanced.
Drawings
FIG. 1 is a flow chart of a data set enhancement method for deep learning assessment of vascular network developmental levels according to the present invention.
Fig. 2 is a specific flow of preprocessing an original picture.
FIG. 3 is a detailed flow chart of data amplification.
FIG. 4 is a specific flow of selecting an amplification data generation dataset.
Detailed Description
For the purpose of better explaining the present invention and to facilitate understanding, the present invention will be described in detail by way of specific embodiments with reference to the accompanying drawings.
Referring to fig. 1 to 4, a data set enhancement method for deep learning assessment of the developmental level of a vascular network includes the steps of:
e. preprocessing an original image;
f. performing data amplification on the preprocessed image;
g. dividing N rectangular regions with equal area in the space range of an original image;
h. and randomly selecting the same proportion of pictures in each area to generate a data set.
Further, in the step a, the pretreatment comprises the following steps:
a1 removing the area of invalid information around the image;
a2, zooming the images generated by different magnifying glass multiples to reach the same scale standard;
a3 binarizes the image to remove noise point interference.
Invalid regions and noise points in the original blood vessel network image can be effectively reduced through the steps.
Still further, in the step b, the data amplification comprises the following steps:
b1 setting a picture coverage threshold E, wherein the threshold is set to avoid the overlarge overlapping area of the amplified pictures;
b2 uniform random spotting is performed on the image after pre-processing. And (3) generating a picture by taking the uniformly and randomly acquired coordinate point as a central point of the amplification picture, calculating the coverage rate of the picture and each existing picture, assuming that the coordinate of the uniformly and randomly acquired point is (X, Y), generating a square picture by taking the point as the central point of the picture, wherein the side length of the picture is C, the coordinate of the other point is (X ', Y'), the side length of the picture is C, and the overlapping area of the two pictures is as follows:
the coverage of these two graphs was calculated as:
b3, if the coverage rate calculated by the picture generated by uniformly and randomly taking the point and the existing picture is lower than the coverage rate threshold value E, the picture generated by the point as the picture center point meets the requirement and is stored, otherwise, the random and uniform point taking is carried out again, and the operation is repeated until the preprocessed picture can not take the coordinate point meeting the condition;
b4 rotates and mirrors the image saved in b3 by a certain angle to generate new data, generates a new data every time rotating by an angle, and generates a new data every time mirroring is performed similarly.
Through the steps, the data volume can be effectively increased, and the generated new data can avoid overlarge overlapped areas among the pictures, so that the over-fitting condition is avoided.
Further, in the step c, the area division includes the following steps:
the division of the c1 area depends on the size of the picture after pretreatment and the size of the picture after amplification, and the picture after pretreatment is divided into N rectangular areas with equal area;
c2, if the size of the preprocessed picture is far larger than that of the data amplified picture, properly increasing the divided rectangular areas; otherwise, the divided rectangular area is reduced appropriately. The method for dividing the rectangular regions is various, the rectangular regions can be divided in a way of being parallel to the length or width of the picture, and the rectangular regions can also be divided into grids, but each rectangular region with the same area can be ensured to have enough central points of the picture to be amplified, so that the picture can be conveniently selected.
The reference of the selection area is provided for generating a uniform data set for the subsequent partition selection through the steps.
In the step d, generating the data set includes the following steps:
d1 setting a ratio, G%, which is used to determine the number of amplifications data for each partition into the training set;
d2 step c the preprocessed image has been divided into N rectangular regions of equal area, denoted A1,A2……ANShowing the divided areas, if the central coordinate point of F pictures falls on A1Region, then we are at A1Randomly selecting [ F X G%]([]Representing to get a whole) of the amplified pictures generated by the central coordinate point to a training set, and putting the rest pictures into a testing set; to A2…ANThe region repeats the above operation.
Through the steps, the generated data set can be distributed more uniformly, and more characteristic information can be learned by the model.
In the embodiment of the invention, a data volume amplification and partition map selection method is utilized, and after data amplification is carried out on an original image, equal-proportion quantity map selection of each area is carried out, so that a balanced data set is generated.
For example, in the present embodiment, since the picture size of the blood vessel network is too large (1920 × 1080, the method of the present invention is suitable for enhancing the data set of the blood vessel network with various sizes), and the number of the pictures is small, the training is not facilitated, and the overfitting situation is caused. To facilitate the training of the model, we need to perform data amplification on it, so that its data set is cut to the appropriate size and added to a certain number for model training.
Referring to fig. 1, fig. 1 is a schematic flow chart illustrating a method for generating a vascular network data set based on deep learning according to the present invention, where the method of the present embodiment includes the following steps:
the original picture is accompanied by interference such as invalid regions and blurred blood vessels, and the steps of preprocessing the original picture in this embodiment are as follows:
some invalid regions without feature information are removed. In this data set, the original image size is 1920 × 1080, the left and right sides of the original image have black invalid regions, and therefore the original image needs to be cropped, and the cropped picture size is 1320 × 1080.
If the image is generated by magnifying with magnifying glasses with different magnification factors, the image needs to be zoomed to reach the same size standard, and the interference of the external factors is reduced. In this embodiment, since the blood vessel network in the stage of 3 days and the blood vessel network in the stage of 4 days are generated by microscope experiments with different magnifications, the image in the stage of 3 days needs to be entirely zoomed so as to have the same size as the blood vessel network in the stage of 4 days, thereby reducing the influence of external factors.
And image processing software is used for carrying out binarization operation, so that the interference of noise points and different illumination conditions on the picture is reduced.
Referring to fig. 2, fig. 2 is a specific flow of preprocessing an original picture.
And performing data amplification on the preprocessed pictures. Due to the particularity of the blood vessel network image, data amplification is generally performed by rotating by a proper angle, mirroring and the like. In order to solve the problem of how to take points during data amplification, in this embodiment, a picture coverage threshold E is first set, and the threshold of the picture repetition rate is set so as to keep a certain space between pictures, avoid excessive overlapping, and keep the overlapped part of the generated pictures below the coverage threshold, thereby ensuring the validity of data.
For example, the pre-processed picture is 1320 × 1080, the size of the augmented picture is 200 × 200, a picture coverage threshold E is set to 50%, it is assumed that the existing coordinate points are (150 ), and the uniformly randomly acquired coordinate points are (200 ). When there is an overlap between the two images, the overlapping area of the two images is calculated to be S2500, and the coverage of the two images is calculated to be D25%. And repeating the operation, if the calculated coverage rate of each existing picture and the pictures generated by the point (200 ) is less than the threshold value, taking the point (200 ) as the picture generated by the center point of the picture to meet the requirement, storing the picture, and if not, uniformly and randomly taking the points again until all the points meeting the condition are taken. This can avoid the data expansion and the overlapping area between the pictures is too large.
Referring to fig. 3, fig. 3 is a process of generating new data by performing data amplification on a picture after a certain preprocessing.
The number of data sets can be effectively increased through the steps, and then the problem of how to select the generated data sets of the pictures is solved.
The data set includes a training set and a test set. If we randomly select pictures in the amplified data set to generate a training set and a testing set, it may cause that too much amplified data is obtained in the left half of some preprocessed pictures in the training set, and too little amplified data is obtained in the right half of the preprocessed pictures in the training set, so that the selection of the training set is not balanced enough. In order to solve this problem, in this embodiment, the preprocessed pictures are divided into rectangular regions with four grids, and in each region, the same proportion of the number of pictures are selected for generating the data set. Assuming that the ratio is 70%, after dividing the preprocessed image into 4 regions, the coordinates of the center point of 10 amplified pictures fall in the 1 st region, we need to randomly select 7 (10 × 70%) pictures in the region to the training set, select the rest pictures to the test set, and repeat the above operations for the other regions. Therefore, the four areas divided by each preprocessed picture have a considerable amount of amplification data in the training set and the test set, and the comprehensiveness of the data set is ensured.
Referring to fig. 4, fig. 4 shows a process of generating a data set for the amplification data of a certain preprocessed picture.
Through the steps, the number of the data sets can be effectively increased, excessive overlapping between the data is avoided, the balance of the samples can be guaranteed, and the training effect of the model is improved.
Claims (5)
1. A data set enhancement method for deep learning assessment of vascular network developmental levels, the method comprising the steps of:
a. preprocessing an original image;
b. performing data amplification on the preprocessed image; the data amplification comprises the following steps:
b1 setting a picture coverage threshold E, wherein the threshold is set to avoid the overlarge overlapping area of the amplified pictures;
b2 performing uniform random point extraction in the preprocessed image, generating an image by taking the uniformly randomly extracted coordinate point as the central point of the amplification image, calculating the coverage rate of the image and each existing image, assuming that the coordinate of the uniform random point extraction is (X, Y), generating a square image by taking the point as the central point of the image, the side length of the image is C, the coordinate of the other point is (X ', Y'), the side length of the image is C, and the overlapping area of the two images is:
the coverage of these two graphs was calculated as:
b3, if the coverage rate calculated by the picture generated by uniformly and randomly taking the point and the existing picture is lower than the coverage rate threshold value E, the picture generated by the point as the picture center point meets the requirement and is stored, otherwise, the random and uniform point taking is carried out again, and the operation is repeated until the preprocessed picture can not take the coordinate point meeting the condition;
b4 rotating and mirroring the image stored in b3 by a certain angle to generate new data, generating a new data every time of rotating by an angle, and generating a new data every time of mirroring in the same way;
c. dividing N rectangular regions with equal area in the space range of an original image;
d. and randomly selecting the same proportion of pictures in each area to generate a data set.
2. The data set enhancement method for deep learning and assessing the developmental level of a vascular network according to claim 1, wherein the preprocessing in the step a comprises the steps of:
a1 removing the area of invalid information around the image;
a2, zooming the images generated by different magnifying glass multiples to reach the same scale standard;
a3 binarizes the image to remove noise point interference.
3. The data set enhancement method for deep learning assessment of the developmental level of vascular networks according to claim 1 or 2, wherein in step c, the region division comprises the following steps:
the division of the c1 area depends on the size of the picture after pretreatment and the size of the picture after amplification, and the picture after pretreatment is divided into N rectangular areas with equal area;
c2, if the size of the preprocessed picture is far larger than that of the data amplified picture, increasing the divided rectangular areas; otherwise, the divided rectangular area is reduced.
4. The data set enhancement method for deep learning and assessing the developmental level of a vascular network according to claim 3, wherein in the step c2, the rectangular region is divided by: the division or the grid division is carried out parallel to the length or the width of the picture, so that the central point of each rectangular region with the same area is ensured to be sufficiently enlarged, and the picture can be conveniently selected.
5. A data set enhancement method for deep learning assessment of the developmental level of vascular networks according to claim 1 or 2, wherein in step d, generating a data set comprises the steps of:
d1 setting a ratio, G%, which is used to determine the number of amplifications data for each partition into the training set;
d2 step c the preprocessed image has been divided into N rectangular regions of equal area, denoted A1,A2……ANShowing the divided areas, if the central coordinate point of F pictures falls on A1Region, then we are at A1Randomly selecting [ F X G%]Expanding the amplified picture generated by the central coordinate point into a training set, and putting the rest pictures into a testing set; to A2…ANThe region repeats the above operation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810261472.0A CN108596840B (en) | 2018-03-28 | 2018-03-28 | Data set enhancement method for deep learning evaluation of vascular network development level |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810261472.0A CN108596840B (en) | 2018-03-28 | 2018-03-28 | Data set enhancement method for deep learning evaluation of vascular network development level |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108596840A CN108596840A (en) | 2018-09-28 |
| CN108596840B true CN108596840B (en) | 2021-10-29 |
Family
ID=63623798
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810261472.0A Active CN108596840B (en) | 2018-03-28 | 2018-03-28 | Data set enhancement method for deep learning evaluation of vascular network development level |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108596840B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111783981A (en) * | 2020-06-29 | 2020-10-16 | 百度在线网络技术（北京）有限公司 | Model training method and device, electronic equipment and readable storage medium |
| CN113223671B (en) * | 2021-05-18 | 2022-05-27 | 浙江工业大学 | Microvascular tree generation method based on conditional generation countermeasure network and constraint rule |
| CN114565590B (en) * | 2022-03-03 | 2022-10-25 | 北京安德医智科技有限公司 | Blood vessel data set amplification method and device, electronic device and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106910187A (en) * | 2017-01-13 | 2017-06-30 | 陕西师范大学 | A kind of artificial amplification method of image data set for Bridge Crack detection |
| CN107316289A (en) * | 2017-06-08 | 2017-11-03 | 华中农业大学 | Crop field spike of rice dividing method based on deep learning and super-pixel segmentation |
-
2018
- 2018-03-28 CN CN201810261472.0A patent/CN108596840B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106910187A (en) * | 2017-01-13 | 2017-06-30 | 陕西师范大学 | A kind of artificial amplification method of image data set for Bridge Crack detection |
| CN107316289A (en) * | 2017-06-08 | 2017-11-03 | 华中农业大学 | Crop field spike of rice dividing method based on deep learning and super-pixel segmentation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108596840A (en) | 2018-09-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108596840B (en) | Data set enhancement method for deep learning evaluation of vascular network development level | |
| CN109934826A (en) | A kind of characteristics of image dividing method based on figure convolutional network | |
| CN109978807A (en) | A kind of shadow removal method based on production confrontation network | |
| CN110245697B (en) | Surface contamination detection method, terminal device and storage medium | |
| CN107590512B (en) | The adaptive approach and system of parameter in a kind of template matching | |
| CN105528784B (en) | A kind of method and apparatus of prospect background segmentation | |
| CN109726649B (en) | Remote sensing image cloud detection method and system and electronic equipment | |
| CN112950554B (en) | Lung lobe segmentation optimization method and system based on lung segmentation | |
| CN111951172A (en) | Image optimization method, device, equipment and storage medium | |
| CN111179196B (en) | Multi-resolution depth network image highlight removing method based on divide-and-conquer | |
| JP2012058787A (en) | Information processor and processing method thereof | |
| CN112348831B (en) | Shale SEM image segmentation method based on machine learning | |
| CN116309483A (en) | DDPM-based semi-supervised power transformation equipment characterization defect detection method and system | |
| CN116052105A (en) | Pavement crack identification classification and area calculation method, system, equipment and terminal | |
| CN115909353A (en) | Image binarization processing method and device | |
| CN113160078B (en) | Method, device and equipment for removing rain from traffic vehicle image in rainy day and readable storage medium | |
| CN114266899A (en) | Image target parallel detection method based on multi-core DSP | |
| CN112381845B (en) | Rock core image generation method, model training method and device | |
| Volkmann | An approach to automated particle picking from electron micrographs based on reduced representation templates | |
| CN111914846B (en) | Layout data synthesis method, equipment and storage medium | |
| CN110675337B (en) | Diffusion type traversal method for image noise reduction | |
| CN110930423B (en) | Object edge feature recognition and extraction method | |
| CN108960285B (en) | Classification model generation method, tongue image classification method and tongue image classification device | |
| CN106023105A (en) | Binary image generation method and system for plant leaves | |
| CN115619800A (en) | COVID-19CT image segmentation method and device based on adaptive threshold segmentation |
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 | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20180928 Assignee: Hangzhou Youshu Cloud Travel Information Technology Co.,Ltd. Assignor: JIANG University OF TECHNOLOGY Contract record no.: X2023980054817 Denomination of invention: A Dataset Enhancement Method for Deep Learning Evaluation of Vascular Network Development Level Granted publication date: 20211029 License type: Common License Record date: 20240102 Application publication date: 20180928 Assignee: Hangzhou Tianyin Computer System Engineering Co.,Ltd. Assignor: JIANG University OF TECHNOLOGY Contract record no.: X2023980054814 Denomination of invention: A Dataset Enhancement Method for Deep Learning Evaluation of Vascular Network Development Level Granted publication date: 20211029 License type: Common License Record date: 20240102 |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20180928 Assignee: HANGZHOU YONGGUAN NETWORK TECHNOLOGY CO.,LTD. Assignor: JIANG University OF TECHNOLOGY Contract record no.: X2024980000361 Denomination of invention: A Dataset Enhancement Method for Deep Learning Evaluation of Vascular Network Development Level Granted publication date: 20211029 License type: Common License Record date: 20240109 |