CN109978027A - An a kind of key drawing methods, system and storage medium based on machine learning - Google Patents
An a kind of key drawing methods, system and storage medium based on machine learning Download PDFInfo
- Publication number
- CN109978027A CN109978027A CN201910184547.4A CN201910184547A CN109978027A CN 109978027 A CN109978027 A CN 109978027A CN 201910184547 A CN201910184547 A CN 201910184547A CN 109978027 A CN109978027 A CN 109978027A
- Authority
- CN
- China
- Prior art keywords
- data
- sample
- key drawing
- user
- drawing methods
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/0007—Image acquisition
Abstract
A present invention provides a kind of key drawing methods, system and storage medium based on machine learning, the one key drawing methods include: step 1, receive element, most like data are obtained according to the sample data that the label of the element of acquisition and training set matches, element includes the background picture that user uploads and the descriptive text of editor；Step 2, the classification classified as new data that frequency of occurrence is most in the most like data of setting quantity is selected；Step 3, multiple sample frames for user's selection are calculated；Step 4, the sample frame of user's selection is received；Step 5, the data most like with element are found out, then directly with the sample frame come at figure；Step 6, it charts successfully, the image to complete is presented to the user.The beneficial effects of the present invention are: the present invention can a key at figure, avoid repeating to do figure, mitigate the workload for doing figure, technical effect is good, the popularization and application of value.
Description
Technical field
The present invention relates to technical field of image processing more particularly to a kind of key drawing methods based on machine learning, it is
System and storage medium.
Background technique
With the development of the society, the development of science and technology, software is widely used in people's life and production, but at present
It does that figure producing efficiency is low, does the heavy workload of figure, be unable to satisfy the demand of user.
Summary of the invention
The present invention provides a kind of key drawing methods based on machine learning, include the following steps:
Step 1, element is received, is obtained according to the sample data that the label of the element of acquisition and training set matches most like
Data, element include user upload background picture and editor descriptive text；
Step 2, the classification classified as new data that frequency of occurrence is most in the most like data of setting quantity is selected；
Step 3, multiple sample frames for user's selection are calculated；
Step 4, the sample frame of user's selection is received；
Step 5, the data most like with element are found out, then directly with the sample frame come at figure；
Step 6, it charts successfully, the image to complete is presented to the user.
As a further improvement of the present invention, in the step 2, select frequency of occurrence in 10 most like data most
Classification of the classification as new data.
As a further improvement of the present invention, in the step 2, classification includes pattern, background, color.
As a further improvement of the present invention, in the step 3,10 sample panes for user's selection are calculated
Frame.
As a further improvement of the present invention, in the step 3,
For data
There are 10 samples, dimension is d (p, q) latitude, and d (p, q) is bivector, and the step of constructing KS tree is as follows:
Step A: construction root node corresponds to the hypermatrix region of all samples comprising d (p, q) according to node, wherein
Qn is n-th of component of vector q；
Step B: the first dimension p, q of selection is coordinate, is to cut with the median of value of the samples all in d in the first dimension
Branch, will be divided into two sub-regions according to the corresponding hypermatrix region of node, cutting by by cut-off and with current selected
The vertical hyperplane of reference axis is realized；The left and right child node that depth is 1 is thus generated by root node: inside left child node
Value of all samples in the first dimension be less than cut-off, value of all samples in the first dimension inside right child node is big
In cut-off；
Step C: repeating step B, the node for being j for current depth, thus generates depth by the node that depth is j+1
Degree is the left and right child node of j+1, until two sub-regions do not have to stop in the presence of example, to form the region division of KS tree.
As a further improvement of the present invention, in the step 5, most similar data I=y is found out(i)=Cj value is 1
When as most like data, then directly with the sample frame come at figure, I is the distance to oneself, y(i)For angle amplitude；
Work as y(i)When=Cj is 0, C is all label types of training set, and Cj is the label of j-th of classification.
As a further improvement of the present invention, further include following steps between the step 5 and the step 6:
It, will be in the color of new data, picture style, verbal description type and sample set for the new data of not label
The corresponding feature of data is compared, and the tag along sort of the most like data of sample is then extracted according to neighbouring method.
As a further improvement of the present invention, the step of extracting the tag along sort of the most like data of sample according to neighbouring method is wrapped
It includes:
Input step: input training dataset T inputs example x；
Export step: class y belonging to output instance X；
Include: in output step
It looks for neighbour: according to given distance metric, being found in training set T and k example x nearest neighbour；
It makes and classifies: determining the classification y of x according to categorised decision rule.
A present invention also provides a kind of key drawing formation system based on machine learning, comprising: memory, processor and deposit
The computer program on the memory is stored up, the present invention is realized when the computer program is configured to be called by the processor
The step of described method.
The present invention also provides a kind of computer readable storage medium, the computer-readable recording medium storage has calculating
The step of machine program, the computer program realizes method of the present invention when being configured to be called by processor.
The beneficial effects of the present invention are: the present invention can a key at figure, avoid repeating to do figure, mitigate the workload for doing figure,
Technical effect is good, the popularization and application of value.
Detailed description of the invention
Fig. 1 is the schematic diagram of the KS tree of construction.
Fig. 2 is the schematic diagram that Nearest Neighbor Search is carried out using KS tree.
Fig. 3 is interaction diagrams of the invention.
Fig. 4 is system operational flow diagram of the invention.
Specific embodiment
The invention discloses a kind of key drawing methods based on machine learning, include the following steps:
Step S1 receives element, obtains most phase according to the sample data that the label of the element of acquisition and training set matches
As data, element include user upload background picture and editor descriptive text；
Step S2, selects the classification that frequency of occurrence is most in 10 most like data, and classification includes (pattern, background, face
Color), the classification as new data；
Step S3 calculates 10 sample frames for user's selection, it is available to be equivalent to 10 kinds of patterns；
Step S4 receives the sample frame of user's selection；
Step S5 finds out the data most like with element, then directly with the sample frame come at figure；
Step S6, for the new data of not label, by the color of new data, picture style, verbal description type with
The corresponding feature of sample intensive data is compared, and the classification of the most like data of sample (arest neighbors) is then extracted according to neighbouring method
Label；
Step S7 charts successfully, the image to complete is presented to the user.
In the step S3, for data
There are 10 samples, dimension is d (p, q) latitude, and d (p, q) is bivector, and (a kind of segmentation k dimension data is empty for construction KS tree
Between data structure) the step of it is as follows:
Step A: construction root node (left subspace and right subspace) corresponds to all samples comprising d (p, q) according to node
The hypermatrix region of example, wherein qn is n-th of component of vector q；
Step B: the first dimension p, q of selection is coordinate, is to cut with the median of value of the samples all in d in the first dimension
Branch, will be divided into two sub-regions according to the corresponding hypermatrix region of node, cutting by by cut-off and with current selected
The vertical hyperplane of reference axis is realized；The left and right child node that depth is 1 is thus generated by root node: inside left child node
Value of all samples in the first dimension be less than cut-off, value of all samples in the first dimension inside right child node is big
In cut-off；
Step C: repeating step B, the node for being j for current depth, thus generates depth by the node that depth is j+1
Degree is the left and right child node of j+1, until two sub-regions do not have to stop in the presence of example, to form the region division of KS tree.
For example, for data set
The KS tree of construction is as shown in Figure 1.We will carry out Nearest Neighbor Search using KS tree now, as shown in Fig. 2, wherein S is test specimens
Example, the right node D comprising S is had found with S first, and then to draw circle of the S as the center of circle by point D, then arest neighbors one is scheduled on this
The inside of a circle, then successively returns to root node, see be corresponding region whether with circle intersect, node more therein with currently
The distance of closest approach, and be updated, to find out nearest neighbor point.
In the step S5, most similar data I=y is found out(i)=Cj value data as most like when being 1, then
Directly with the sample frame come at figure, I is the distance to oneself, y(i)For angle amplitude；Work as y(i)When=Cj is 0, C is training set
All label types, Cj be j-th of classification label.
In step s 6, include: according to the step of tag along sort of the neighbouring method extraction most like data of sample
Input step: input training dataset T (picture) inputs example x (text)；
It looks for neighbour: according to given distance metric, being found in training set T and k example x nearest neighbour；
It makes and classifies: determining the classification y of x according to categorised decision rule.
Key drawing of the invention includes a key banner, a key poster, a key active page.
As shown in figure 3, user enters login module by account number cipher, drawing request is then initiated, drawing module is shown
Type of charting (drawing type includes a key banner, a key poster, a key active page) is to user, after user selects drawing type
Start key drawing, starting one key drawing just brings into operation the key drawing methods of the invention based on machine learning, drawing at
After function, figure is returned into user.
As shown in figure 4, the key of the invention based on machine learning that just brings into operation is at figure side when executing a key into figure
Method.
A invention also discloses a kind of key drawing formation system based on machine learning, comprising: memory, processor and deposit
The computer program on the memory is stored up, the present invention is realized when the computer program is configured to be called by the processor
The step of described key drawing methods.
The invention also discloses a kind of computer readable storage medium, the computer-readable recording medium storage has calculating
The step of machine program, the computer program realizes key drawing methods of the present invention when being configured to be called by processor.
The present invention can a key at figure, avoid repeating to do figure, mitigate the workload for doing figure, technical effect is good, the popularization of value
Using.
The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be said that
Specific implementation of the invention is only limited to these instructions.For those of ordinary skill in the art to which the present invention belongs, exist
Under the premise of not departing from present inventive concept, a number of simple deductions or replacements can also be made, all shall be regarded as belonging to of the invention
Protection scope.
Claims (10)
1. a kind of key drawing methods based on machine learning, which comprises the steps of:
Step 1, element is received, most like number is obtained according to the sample data that the label of the element of acquisition and training set matches
According to element includes the background picture that user uploads and the descriptive text of editor；
Step 2, the classification classified as new data that frequency of occurrence is most in the most like data of setting quantity is selected；
Step 3, multiple sample frames for user's selection are calculated；
Step 4, the sample frame of user's selection is received；
Step 5, the data most like with element are found out, then directly with the sample frame come at figure；
Step 6, it charts successfully, the image to complete is presented to the user.
2. key drawing methods according to claim 1, it is characterised in that: in the step 2, selection 10 is most like
Classification of the most classification of frequency of occurrence as new data in data.
3. key drawing methods according to claim 1, it is characterised in that: in the step 2, classification include pattern,
Background, color.
4. key drawing methods according to claim 1, it is characterised in that: in the step 3, calculate for user
10 sample frames of selection.
5. key drawing methods according to claim 1, it is characterised in that: in the step 3, for data
There are 10 samples, dimension is d (p, q) latitude, and d (p, q) is bivector, and the step of constructing KS tree is as follows:
Step A: construction root node corresponds to the hypermatrix region of all samples comprising d (p, q) according to node, and wherein qn is
N-th of component of vector q；
Step B: the first dimension p, q of selection is coordinate, using the median of value of the samples all in d in the first dimension as cutting
Point, will be divided into two sub-regions according to the corresponding hypermatrix region of node, cutting by by cut-off and with the seat of current selected
The vertical hyperplane of parameter is realized；The left and right child node that depth is 1 is thus generated by root node: inside left child node
Value of all samples in the first dimension is less than cut-off, and value of all samples in the first dimension inside right child node is greater than
Cut-off；
Step C: repeating step B, and the node for being j for current depth, thus generating depth by the node that depth is j+1 is
The left and right child node of j+1, until two sub-regions do not have to stop in the presence of example, to form the region division of KS tree.
6. key drawing methods according to claim 1, it is characterised in that: in the step 5, find out most similar number
According to I=y(i)As most like data when=Cj value is 1, then directly with the sample frame come at figure, I be to oneself away from
From y(i)For angle amplitude；Work as y(i)When=Cj is 0, C is all label types of training set, and Cj is the label of j-th of classification.
7. key drawing methods according to claim 1, which is characterized in that between the step 5 and the step 6 also
Include the following steps:
For the new data of not label, by the color of new data, picture style, verbal description type and sample intensive data
Corresponding feature is compared, and the tag along sort of the most like data of sample is then extracted according to neighbouring method.
8. key drawing methods according to claim 7, which is characterized in that extract the most like data of sample according to neighbouring method
Tag along sort the step of include:
Input step: input training dataset T inputs example x；
Export step: class y belonging to output instance X；
Include: in output step
It looks for neighbour: according to given distance metric, being found in training set T and k example x nearest neighbour；
It makes and classifies: determining the classification y of x according to categorised decision rule.
9. an a kind of key drawing formation system based on machine learning characterized by comprising memory, processor and be stored in
Computer program on the memory realizes claim 1- when the computer program is configured to be called by the processor
The step of method described in any one of 8.
10. a kind of computer readable storage medium, which is characterized in that the computer-readable recording medium storage has computer journey
The step of sequence, the computer program realizes method of any of claims 1-8 when being configured to be called by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910184547.4A CN109978027A (en) | 2019-03-12 | 2019-03-12 | An a kind of key drawing methods, system and storage medium based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910184547.4A CN109978027A (en) | 2019-03-12 | 2019-03-12 | An a kind of key drawing methods, system and storage medium based on machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109978027A true CN109978027A (en) | 2019-07-05 |
Family
ID=67078579
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910184547.4A Pending CN109978027A (en) | 2019-03-12 | 2019-03-12 | An a kind of key drawing methods, system and storage medium based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109978027A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103885951A (en) * | 2012-12-19 | 2014-06-25 | 阿里巴巴集团控股有限公司 | Graphics and text information releasing and generating method and graphics and text information releasing and generating device |
| CN107301566A (en) * | 2017-06-19 | 2017-10-27 | 郑州航空工业管理学院 | A kind of advertisement design multimedia information system |
| CN108416827A (en) * | 2018-02-11 | 2018-08-17 | 李荣陆 | A kind of planar design autoplacement device and method analyzed based on data-driven and material |
| CN108694210A (en) * | 2017-04-11 | 2018-10-23 | 阿里巴巴集团控股有限公司 | Template generation method and device |
| CN108694602A (en) * | 2017-04-11 | 2018-10-23 | 阿里巴巴集团控股有限公司 | Promotional literature generation method and device |
| CN109242570A (en) * | 2018-09-17 | 2019-01-18 | 泰兴市唯艺传媒广告有限公司 | A kind of advertisement interactive design system |
| CN109447683A (en) * | 2018-09-25 | 2019-03-08 | 泰兴市唯艺传媒广告有限公司 | A kind of human-computer exchange formula advertisement design system |
-
2019
- 2019-03-12 CN CN201910184547.4A patent/CN109978027A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103885951A (en) * | 2012-12-19 | 2014-06-25 | 阿里巴巴集团控股有限公司 | Graphics and text information releasing and generating method and graphics and text information releasing and generating device |
| CN108694210A (en) * | 2017-04-11 | 2018-10-23 | 阿里巴巴集团控股有限公司 | Template generation method and device |
| CN108694602A (en) * | 2017-04-11 | 2018-10-23 | 阿里巴巴集团控股有限公司 | Promotional literature generation method and device |
| CN107301566A (en) * | 2017-06-19 | 2017-10-27 | 郑州航空工业管理学院 | A kind of advertisement design multimedia information system |
| CN108416827A (en) * | 2018-02-11 | 2018-08-17 | 李荣陆 | A kind of planar design autoplacement device and method analyzed based on data-driven and material |
| CN109242570A (en) * | 2018-09-17 | 2019-01-18 | 泰兴市唯艺传媒广告有限公司 | A kind of advertisement interactive design system |
| CN109447683A (en) * | 2018-09-25 | 2019-03-08 | 泰兴市唯艺传媒广告有限公司 | A kind of human-computer exchange formula advertisement design system |
Non-Patent Citations (1)
| Title |
|---|
| 杜志民: "基于局部不变特征的图像检索系统的研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10885323B2 (en) | Digital image-based document digitization using a graph model | |
| CN104573130B (en) | The entity resolution method and device calculated based on colony | |
| CN103812872B (en) | A kind of network navy behavioral value method and system based on mixing Di Li Cray process | |
| CN108446964B (en) | User recommendation method based on mobile traffic DPI data | |
| WO2021109464A1 (en) | Personalized teaching resource recommendation method for large-scale users | |
| CN106815307A (en) | Public Culture knowledge mapping platform and its use method | |
| CN103106262B (en) | The method and apparatus that document classification, supporting vector machine model generate | |
| US20210049478A1 (en) | Feature relationship recommendation method, apparatus, computing device, and storage medium | |
| CN106557558A (en) | A kind of data analysing method and device | |
| JP6428795B2 (en) | Model generation method, word weighting method, model generation device, word weighting device, device, computer program, and computer storage medium | |
| CN110287329A (en) | A kind of electric business classification attribute excavation method based on commodity text classification | |
| CN106445915A (en) | New word discovery method and device | |
| CN107133854A (en) | Information recommendation method and device | |
| CN109308324A (en) | A kind of image search method and system based on hand drawing style recommendation | |
| CN110647645A (en) | Attack image retrieval method based on general disturbance | |
| CN110766272A (en) | Power business collaborative classification method and system based on ID3 decision tree algorithm | |
| CN116049379A (en) | Knowledge recommendation method, knowledge recommendation device, electronic equipment and storage medium | |
| CN104008177A (en) | Method and system for rule base structure optimization and generation facing image semantic annotation | |
| CN105528432B (en) | A kind of digital resource hot spot generation method and device | |
| CN107133321B (en) | Method and device for analyzing search characteristics of page | |
| CN111708919B (en) | Big data processing method and system | |
| CN107391650A (en) | A kind of structuring method for splitting of document, apparatus and system | |
| JP2020502710A (en) | Web page main image recognition method and apparatus | |
| CN108830302B (en) | Image classification method, training method, classification prediction method and related device | |
| CN111831819A (en) | Text updating method and device |
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 |