KR20230017628A - Platform system for extracting image and for learning image - Google Patents

Abstract

The present invention provides an image extraction and training platform for artificial intelligence learning. The image extraction and training platform for artificial intelligence learning includes: a video data storage unit containing a plurality of video data including images; a video data extraction unit for selecting a training target image from among the plurality of video data including the images from the video data storage unit and extracting the selected training target image; and an image training unit for presetting reference information corresponding to the plurality of video data including the images and repeatedly training the extracted training target image to reach the accuracy of the reference information. Accordingly, image accuracy can be improved to the target quality.

Description

Image extraction and learning platform for artificial intelligence learning {Platform system for extracting image and for learning image}

The present invention relates to an image extraction and learning platform for artificial intelligence learning, and more particularly, extracts only a target image from various image data held by personal smart devices, the Internet, and public institutions, and based on the extracted image, the learning data It is about an image extraction and learning platform for artificial intelligence learning that can improve the accuracy of images to a target quality by learning.

Artificial intelligence (AI) usually refers to intelligence created by machines.

It is a field of computer engineering and information technology that studies ways to enable computers to think, learn, etc. that can be done with human intelligence. .

Artificial intelligence is constantly being researched and developed, from image intelligence to voice and text intelligence, and currently, research and development on video image intelligence is being conducted, and it is growing rapidly, and the industrial ripple effect of video image intelligence is very large.

It is important for recent artificial intelligence to prevent errors by self-learning based on experience, and to process the intended processing task with a low error rate.

Accordingly, even if the manager does not input the learning data one by one, the information on the desired image is not simply obtained in the process of acquiring, but is learned and stored to achieve the required accuracy, and errors are eliminated in processing the image in the future. It is important to reduce and achieve accurate processing.

Republic of Korea Patent Registration No. 10-2015939 (registration date: August 23, 2019)

The present invention has been made to solve the above problems, and the present invention extracts only the desired image from various image data held by personal smart devices, the Internet, and public institutions, and learns through pre-learning data based on the extracted image. Its purpose is to provide an image extraction and learning platform for artificial intelligence learning that can improve the accuracy of images to target quality.

In addition, the present invention determines an image to be learned, extracts a learning image that matches the image from various video data, and simultaneously extracts text data and converts it into an image, thereby creating an image that embodies a character including a video. Its purpose is to provide an image extraction and learning platform for artificial intelligence learning that can be supplemented and trained to achieve the required accuracy.

The objects of the present invention are not limited to the above-mentioned objects, and other objects and advantages of the present invention not mentioned above can be understood by the following description and will be more clearly understood by the examples of the present invention. It will also be readily apparent that the objects and advantages of the present invention may be realized by means of the instrumentalities and combinations indicated in the claims.

In order to achieve the above objects, the present invention provides an image extraction and learning platform for artificial intelligence learning.

The image extraction and learning platform for artificial intelligence learning includes a video data storage unit in which a plurality of video data including images exist; a video data extraction unit for selecting a learning target image from among a plurality of video data including the image from the video data storage unit and extracting the selected learning target image; and an image learning unit in which reference information corresponding to a plurality of video data including the image is preset, and the accuracy of the extracted learning target image is repeatedly learned to reach the accuracy of the reference information.

Here, the image extraction unit,

It is preferable to selectively extract some or all of the learning target images.

And the video data extraction unit,

a learning target image setting unit configured to set the learning target image;

a learning target image selector for selecting an image matching the set learning target image from the plurality of video data;

It is preferable to include a learning target image extraction unit for extracting the selected learning target image from the plurality of image data.

In addition, the video data extraction unit,

Transmitting image extraction request information to the image data storage unit;

When receiving image extraction permission information according to the image extraction request information from the image data storage unit,

Preferably, the learning target image is extracted.

In addition, the image learning unit,

a reference information storage unit for storing the reference information;

an image selector for selecting learning criterion information matching the extracted learning target image from the criterion information;

It is preferable to include an image correction unit for learning through repetitive correction so that the accuracy of the extracted learning target image matches the accuracy of the learning reference information.

In addition, the image correction unit,

It is preferable to set a plurality of the learning criterion information and sequentially correct them.

The image correction unit,

The accuracy of the extracted learning target image is calculated as correction accuracy through the difference between the accuracy of the learning reference information,

sequentially selecting and correcting a plurality of learning criteria information so that the correction accuracy is satisfied to reach the accuracy of the learning criteria information;

After correction, if the correction accuracy does not reach the accuracy of the learning reference information,

Further calculating additional accuracy of the learning criterion information;

Searching for a correction image that can satisfy the additional accuracy from the image providing unit;

Replace the searched correction image with the learning reference information,

extracting difference information that differs from the accuracy of the learning reference information from the extracted learning target image;

Preferably, the difference information is included in replacing the learning criterion information.

In particular, the image correction unit extracts text information from the image data extraction unit.

Here, the text information is extracted, and the image correction unit extracts text information capable of forming the shape of the learning target image.

And the image compensator forms a virtual image based on the extracted character information.

Then, the virtual image is corrected to reach the accuracy of the learning reference information.

The image correction unit stores a virtual image corrected to reach the accuracy of the learning reference information and converts the virtual image into text information.

Subsequently, the image correction unit may store the converted text information.

In accordance with the above tasks, the present invention extracts only the desired image from various image data held by personal smart devices, the Internet, and public institutions, and learns the pre-learning data based on the extracted image to achieve the target quality for accuracy of the image. It has an effect that can be improved.

In addition, the present invention determines an image to be learned, extracts a learning image that matches the image from various video data, and simultaneously extracts text data and converts it into an image, thereby creating an image that embodies a character including a video. It has the effect of being able to train the image to achieve the required accuracy by complementing it.

In addition to the effects described above, specific effects of the present invention will be described together while explaining specific details for carrying out the present invention.

1 is a block diagram showing the configuration of an image extraction and learning platform for artificial intelligence learning according to the present invention.
2 is a block diagram showing a video data extraction unit according to the present invention.
3 is a flowchart showing the driving of the video data extraction unit according to the present invention.
4 is a block diagram showing an image learning unit according to the present invention.
5 is a flowchart showing the driving of an image learning unit according to the present invention.
6 is a flowchart showing another example of image learning according to the present invention.

The above objects, features and advantages will be described later in detail with reference to the accompanying drawings, and accordingly, those skilled in the art to which the present invention belongs will be able to easily implement the technical spirit of the present invention. In describing the present invention, if it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted. Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings. In the drawings, the same reference numerals are used to indicate the same or similar components.

Hereinafter, the arrangement of an arbitrary element on the "upper (or lower)" or "upper (or lower)" of a component means that an arbitrary element is placed in contact with the upper (or lower) surface of the component. In addition, it may mean that other components may be interposed between the component and any component disposed on (or under) the component.

In addition, when a component is described as "connected", "coupled" or "connected" to another component, the components may be directly connected or connected to each other, but other components may be "interposed" between each component. ", or each component may be "connected", "coupled" or "connected" through other components.

Hereinafter, an image extraction and learning platform for artificial intelligence learning according to the present invention will be described with reference to the accompanying drawings.

1 is a block diagram showing the configuration of an image extraction and learning platform for artificial intelligence learning according to the present invention.

Referring to FIG. 1 , an image extraction and learning platform for artificial intelligence learning according to the present invention includes an image data storage unit 100, an image data extraction unit 200, and an image learning unit 300.

A plurality of video data exists in the video data storage unit 100 according to the present invention.

The video data storage unit 100 may include a device such as a personal smart phone or a data server.

Images stored in the video data storage unit 100 may be video data such as photos and videos. Objects included in the photos and videos may include all objects in daily life, such as people and objects.

The video data storage unit 100 and the video data extraction unit can be applied to both wireless or wired communication, so that information such as images and images can be transmitted through electrical signals.

The video data extractor 200 according to the present invention has a function of selecting a learning target image from among a plurality of video data including the image from the video data storage unit 100 and extracting the selected learning target image. can have

In addition, the image data extraction unit 200 is connected to the image learning unit 300 through wireless communication or wired communication.

The image learning unit 300 may have a function of repeatedly learning the accuracy of the extracted learning target image to reach the accuracy of the reference information, in which reference information corresponding to a plurality of video data including images is preset. .

Next, the configuration and functions of the video data storage unit, the video data extraction unit, and the image learning unit according to the present invention as described above will be described.

2 is a block diagram showing a video data extraction unit according to the present invention. 3 is a flowchart showing the driving of the video data extraction unit according to the present invention.

Referring to FIGS. 2 and 3, the video data extractor 200 sets the learning target image setting unit 210 to set the learning target image, and the image matching the set learning target image to the plurality of video data. It includes a learning target image selection unit 220 that selects from, and a learning target image extraction unit 230 that extracts the selected learning target image from the plurality of video data.

Subsequently, when the video data extractor 200 transmits image extraction request information to the video data storage unit 100 and receives image extraction permission information according to the image extraction request information from the video data storage unit , the learning target image can be extracted.

That is, referring to FIG. 3 , the video data extraction unit 200 according to the present invention transmits image extraction request information to the video data storage unit 100 to determine whether or not to extract an image.

Subsequently, the video data storage unit 100 may receive image extraction request information, and transmits a signal regarding image extraction permission to the video data extraction unit 200 when the image extraction information is permitted.

Next, the learning target image setting unit 210 sets the learning target image. The learning target image may be part or all of the image. Preferably, it is an image whose accuracy or resolution is below a certain level, and it may be an image for which accuracy or resolution is required to be corrected to a certain level or higher.

The learning target image selector 220 selects or determines a set learning target image.

Next, the learning target image extraction unit 230 extracts the selected learning target image from images or videos stored in the image data storage unit 100 .

Accordingly, in the present invention, in order to improve accuracy, a certain learning target image may be extracted from the video data storage unit 100 and transmitted to the image learning unit 300.

4 is a block diagram showing an image learning unit according to the present invention. 5 is a flowchart showing the driving of an image learning unit according to the present invention.

Referring to FIGS. 4 and 5 , the image learning unit 300 according to the present invention transmits the reference information storage unit 310 in which the reference information is stored, and the learning reference information matching the extracted training target image to the reference information. It includes an image selection unit 320 that selects from the images, and an image correction unit 330 that learns through iterative correction so that the accuracy of the extracted learning target image matches the accuracy of the learning reference information.

In addition, the image correction unit 330 according to the present invention may set the learning criterion information to a plurality and sequentially correct them.

In addition, the image correction unit 330 calculates the correction accuracy through the difference between the accuracy of the extracted learning target image and the accuracy of the learning reference information.

Subsequently, the image correction unit 330 sequentially selects and corrects a plurality of learning reference information so that the calibration accuracy is satisfied so as to reach the accuracy of the learning reference information.

After the correction, the image correction unit 330 additionally calculates additional accuracy of the learning standard information when the correction accuracy does not reach the accuracy of the learning standard information.

The image correction unit 330 may search the image data storage unit 100 for a correction image that can satisfy the additional accuracy.

Subsequently, the image correction unit 330 may replace the searched correction image with the learning reference information.

Here, the image correction unit according to the present invention replaces the searched corrected image with the learning reference information, and extracts difference information that makes a difference in accuracy of the learning reference information from the extracted learning target image.

And, the image correction unit 330 may include the difference information in replacing the learning criterion information.

That is, in the present invention, after separately extracting learning reference information and information having a tooth from a learning target image, it is applied to replace the learning reference information to improve accuracy to a certain level or more when comparing other learning target images later.

6 is a flowchart showing another example of image learning according to the present invention.

Referring to FIG. 6 , the image correction unit 330 according to the present invention extracts text information from the video data extraction unit 230 .

Here, while extracting the text information, the image correction unit 330 extracts text information capable of forming the shape of the learning target image.

And the image compensator 330 forms a virtual image based on the extracted character information.

Then, the virtual image is corrected to reach the accuracy of the learning reference information.

The image correction unit 330 stores the corrected virtual image to reach the accuracy of the learning reference information and converts the virtual image into text information.

Subsequently, the image correction unit 330 may store the converted text information.

According to the above configuration and operation, the present invention extracts only the target image from various image data held by personal smart devices, the Internet, and public institutions, and aims at the accuracy of the image by learning the pre-learning data based on the extracted image. quality can be improved.

In addition, the present invention determines an image to be learned, extracts a learning image that matches the image from various video data, and simultaneously extracts text data and converts it into an image, thereby creating an image that embodies a character including a video. By supplementing, the image can be trained to achieve the required accuracy.

The above-described present invention, since various substitutions, modifications, and changes are possible to those skilled in the art without departing from the technical spirit of the present invention, the above-described embodiments and accompanying drawings is not limited by

100: image data storage unit
200: video data extraction unit
210: learning target image setting unit
220: learning target image selection unit
230: learning target image extraction unit
300: image learning unit
310: reference image storage unit
320: image selection unit
330: image correction unit

Claims (7)

An image data storage unit in which a plurality of image data including images exist;
a video data extraction unit for selecting a learning target image from among a plurality of video data including the image from the video data storage unit and extracting the selected learning target image; and,
an image learning unit that sets reference information corresponding to a plurality of video data including the image in advance and repeatedly learns the accuracy of the extracted learning target image to reach the accuracy of the reference information;
Image extraction and learning platform for artificial intelligence learning, characterized in that it comprises a.
According to claim 1,
The image extraction unit,
An image extraction and learning platform for artificial intelligence learning, characterized in that for selectively extracting some or all of the learning target images.
According to claim 1,
The video data extraction unit,
a learning target image setting unit configured to set the learning target image;
a learning target image selector for selecting an image matching the set learning target image from the plurality of video data;
An image extraction and learning platform for artificial intelligence learning, characterized in that it comprises a learning target image extractor for extracting the selected learning target image from the plurality of image data.
According to claim 3,
The video data extraction unit,
Transmitting image extraction request information to the image data storage unit;
When receiving image extraction permission information according to the image extraction request information from the image data storage unit,
An image extraction and learning platform for artificial intelligence learning, characterized in that for extracting the learning target image.
According to claim 3,
The image learning unit,
a reference information storage unit for storing the reference information;
an image selector for selecting learning criterion information matching the extracted learning target image from the criterion information;
An image extraction and learning platform for artificial intelligence learning, characterized in that it comprises an image correction unit for learning through repetitive correction the accuracy of the extracted learning target image to match the accuracy of the learning reference information.
According to claim 5,
The image correction unit,
An image extraction and learning platform for artificial intelligence learning, characterized in that by setting a plurality of learning reference information and sequentially correcting them.
According to claim 5,
The image correction unit,
The accuracy of the extracted learning target image is calculated as correction accuracy through the difference between the accuracy of the learning reference information,
sequentially selecting and correcting a plurality of learning criteria information so that the correction accuracy is satisfied to reach the accuracy of the learning criteria information;
After correction, if the correction accuracy does not reach the accuracy of the learning reference information,
Further calculating additional accuracy of the learning criterion information;
Searching for a correction image that can satisfy the additional accuracy from the image providing unit;
Replace the searched correction image with the learning reference information,
extracting difference information that differs from the accuracy of the learning reference information from the extracted learning target image;
Image extraction and learning platform for artificial intelligence learning, characterized in that included in replacing the difference information with the learning reference information.

Images