KR20190054993A - Learning apparatus based on game data and method for the same - Google Patents

Abstract

Disclosed are a learning apparatus based on game data and a method thereof. According to one embodiment of the present invention, the learning apparatus based on game data may comprise: a game data-based image information generating unit constituting a three-dimensional object included in the game data, controlling an environment in which the three-dimensional object is configured, and generating image information including an image photographing the three-dimensional object and environment information indicating an environment in which the three-dimensional object is configured; and a learning processing unit constructing a learning model based on the game data by using the image information provided by the image information generating unit.

Description

TECHNICAL FIELD [0001] The present invention relates to a learning apparatus based on game data,

This disclosure relates to machine learning techniques, and more particularly to methods and apparatus for enhancing data used in machine learning and learning using the same.

Machine learning is a technique that learns universal knowledge from data and classifies data that is not used in learning based on that knowledge.

Recently, machine learning technology based on deep neural network structure has been studied variously in machine learning technology. Generally, a deep neural network is composed of an artificial neural network (ANN) structure including multiple hidden layers between an input layer and an output layer. Since such a neural network can learn various nonlinear relations including multiple hidden layers, it has a great learning potential as an inherent characteristic, and is attracting attention as an effective method in the big data era.

However, large amounts of learning data are required to learn well the systems with large learning potential. Supervised learning methods, which generally show the highest performance in machine learning, require labeled data during learning, and it is not easy and costly to obtain a large amount of labeled data even in the age of big data . In addition, there is a problem that overfitting occurs when the number of labeled data used as learning data is not large.

It is an object of the present invention to provide a game data base learning apparatus and method for enhancing a large amount of data so as to stably perform machine learning based on a deep neural network structure.

It is another object of the present invention to provide a game data-based learning apparatus and method capable of quickly and efficiently enhancing large-capacity data to be used for machine learning based on a neural network structure.

The technical objects to be achieved by the present disclosure are not limited to the above-mentioned technical subjects, and other technical subjects which are not mentioned are to be clearly understood from the following description to those skilled in the art It will be possible.

According to an aspect of the present disclosure, a learning apparatus based on game data can be provided. The apparatus comprises a 3D object included in the game data, controls an environment in which the 3D object is configured, generates image information including environment information indicating an environment in which the 3D object is photographed and the environment in which the 3D object is configured And a learning processing unit for building a learning model based on game data using the image information provided by the image information generating unit.

The features briefly summarized above for this disclosure are only exemplary aspects of the detailed description of the disclosure which follow, and are not intended to limit the scope of the disclosure.

According to the present disclosure, a game data-based learning apparatus and method for enhancing a large amount of data can be provided so as to stably perform machine learning based on a depth-of-field neural network structure.

According to the present disclosure, a game data-based learning apparatus and method capable of quickly and efficiently enhancing large-capacity data to be used for machine learning based on a deep-layer neural network structure can be provided.

According to the present disclosure, a game data-based learning apparatus and method capable of efficiently enhancing large-capacity data to be used for machine learning based on a deep-layer neural network structure using game data can be provided.

The effects obtainable from the present disclosure are not limited to the effects mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the description below will be.

1 is a block diagram showing a configuration of a game data-based learning apparatus according to an embodiment of the present disclosure.
2 is a block diagram showing a detailed configuration of a learning processing unit provided in a game data-based learning apparatus according to an embodiment of the present disclosure;
FIG. 3 is a diagram illustrating operations of a feature extraction unit and a switching unit included in a game data-based learning apparatus according to an embodiment of the present disclosure.
4 is a diagram illustrating operations of a classifier provided in a game data-based learning apparatus according to an embodiment of the present disclosure.
5 is a flowchart showing a procedure of a game data-based learning method according to an embodiment of the present disclosure.
6 is a block diagram illustrating a computing system for executing a game data based learning method and apparatus in accordance with an embodiment of the present disclosure.

Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings, which will be easily understood by those skilled in the art. However, the present disclosure may be embodied in many different forms and is not limited to the embodiments described herein.

In the following description of the embodiments of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present disclosure rather unclear. Parts not related to the description of the present disclosure in the drawings are omitted, and like parts are denoted by similar reference numerals.

In the present disclosure, when an element is referred to as being "connected", "coupled", or "connected" to another element, it is understood that not only a direct connection relationship but also an indirect connection relationship May also be included. Also, when an element is referred to as " comprising " or " having " another element, it is meant to include not only excluding another element but also another element .

In the present disclosure, the components that are distinguished from each other are intended to clearly illustrate each feature and do not necessarily mean that components are separate. That is, a plurality of components may be integrated into one hardware or software unit, or a single component may be distributed into a plurality of hardware or software units. Thus, unless otherwise noted, such integrated or distributed embodiments are also included within the scope of this disclosure.

In the present disclosure, the components described in the various embodiments are not necessarily essential components, and some may be optional components. Thus, embodiments consisting of a subset of the components described in one embodiment are also included within the scope of the present disclosure. Also, embodiments that include other elements in addition to the elements described in the various embodiments are also included in the scope of the present disclosure.

Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings.

1 is a block diagram showing a configuration of a game data-based learning apparatus according to an embodiment of the present disclosure.

Referring to FIG. 1, the game data-based data learning apparatus may include an image information generation unit 11 based on game data and a learning processing unit 15.

The image information generating unit 11 includes a 3D object included in game data, an environment in which the 3D object is configured, and an image of the 3D object and information indicating an environment in which the 3D object is configured Image information can be generated.

Specifically, the 3D game can be configured to extract / modify a 3D model or the like so as to match the taste of the user (gamer). For example, data constituting a 3D game may include data corresponding to a 3D object constituting the game, that is, mesh data, texture data, and the like. The image information generating unit 11 can extract data corresponding to the 3D object, that is, mesh data, texture data, and the like, from the game data. The image information generating unit 11 may generate 3D objects by rendering data corresponding to 3D objects, that is, mesh data, texture data, and the like.

Furthermore, depending on the environment in which the 3D object is placed, the 3D object may be represented and configured differently. For example, the 3D object may be configured in various ways according to the setting of the direction of shooting the 3D object, the color of the 3D object, the lighting environment around the 3D object, and the like. Based on this, the data corresponding to the 3D object may include information (hereinafter, referred to as 'basic environment information') representing a basic environment for representing the 3D object, and the image information generating unit 11 basically has a basic (Hereinafter, referred to as a 'basic image') in which a 3D object is photographed based on environmental information.

Furthermore, the image information generating unit 11 can provide an environment in which the environment in which the 3D object is placed can be variously modified. For example, the image information generating unit 11 may provide a user interface for receiving environmental information used in constructing a 3D object from a user. Accordingly, the image information generating unit 11 can confirm the environment information input by the user through the user interface, and configure the 3D object reflecting the confirmed environment information. As another example, the image information generating unit 11 may randomly generate environment information used in constructing the 3D object, and may construct a 3D object reflecting the environment information generated randomly.

Hereinafter, the image information generating unit 11 defines deformation environment information in order to distinguish environmental information constructed by variously modifying the environment in which the 3D object is placed, from the above-described basic environmental information. Also, the image of the 3D object based on the deformed environment information is defined as the deformed image.

As described above, the image information generating unit 11 can generate a basic image, a deformed image, and the like, which are obtained by photographing a 3D object. In particular, the image information generating unit 11 generates environment information used in constructing the 3D object, Information and the like can be extracted together and provided to the learning processing unit 15. [

On the other hand, the learning processing unit 15 can build a learning model using the image information provided by the image information generating unit 11.

Here, the image information provided by the image information generating unit 11 may include basic image information and deformation environment information corresponding to the basic image and the deformation image generated based on the game data. In particular, since the deformed image and the deformed environment information are information transformed from the basic image and the basic environment information, they may have predetermined characteristics corresponding to the basic image and the basic environment information. In consideration of this, the learning processing unit 15 can confirm the characteristic relationship between the basic image, the basic environment information, the deformed image and the deformed environment information, and build a learning model for the characteristic relation through learning.

Further, when directly extracting the image extracted from the game data, the task domain for us to solve the problem of reality may be different. Therefore, the learning processing unit 15 may perform the transition learning for the image extracted from the game data To build a learning model.

The specific configuration and operation of the learning processing unit provided in the game data-based learning apparatus according to an embodiment of the present disclosure will be described in detail with reference to FIG. 2 through FIG.

2 is a block diagram showing a detailed configuration of a learning processing unit provided in a game data-based learning apparatus according to an embodiment of the present disclosure;

The learning processing unit 15 may include a feature extracting unit 15a, a switching unit 15b, and a classifying unit 15c.

The feature extraction unit 15a can receive the basic image 301 (see FIG. 3) and the transformed image 302 and can extract the feature of the basic image 301 and the transformed image 302, A basic feature map 311 and a modified feature map 312 can be constructed. Furthermore, a plurality of deformed images may be generated by the image information generating unit 11, and the feature extracting unit 15a may constitute a deformed feature map for each of the plurality of deformed images.

Then, the switching unit 15b learns the relationship between the basic feature map 311 and the deformation feature map 312 through the in-depth neural network, and builds a learning model based on the game data. At this time, the switching unit 15b can use the basic environment information and the deformation environment information described above for learning. Furthermore, the structure of the neural network can be composed of Auto-Associative Neural Network, Restricted Boltzman Machine, and Fully Connected Network.

The switching unit 15b may have a plurality of neural networks, and the same neural networks may be overlapped or different neural networks may be combined when constructing a plurality of neural networks.

On the other hand, the classifying unit 15c can re-learn the learning model based on the game data established by the switching unit 15b, and update the neural network included therein.

More specifically, as described above, the switching unit 15b can construct a learning model corresponding to the basic feature map 311 and the modified feature map 312 extracted from the basic image 301 and the transformed image 302 , And a virtual feature map can be constructed using this learning model. For example, an image to be used as actual learning data to be used in an application field (hereinafter referred to as an 'learning object image') 401 is input to the learning processing unit 15, The unit 15a can construct a feature map (hereinafter referred to as a 'learning feature map') 411 by extracting the feature of the learning target image 401. [ The switching unit 15b can construct a virtual feature map 412 corresponding to the learning subject characteristic map 411 using the pre-built learning model. At this time, the virtual feature map 412 has the same semantic label as the learning feature map 411 because the learning model is applied to the learning feature map 411 to increase data.

The virtual characteristic map 412 constructed in this way can be input to the classifying section 15c and the classifying section 15c can use the virtual characteristic map 412 in which the data enhancement is performed and the learning target characteristic map 411 , And perform fine tuning on the learning model based on the game data.

5 is a flowchart showing a procedure of a game data-based learning method according to an embodiment of the present disclosure.

The game data-based learning method according to an embodiment of the present disclosure can be performed by the above-described game data-based learning apparatus.

First, the 3D game can be configured to extract / modify a 3D model or the like so as to match the taste of the user (gamer). For example, data constituting a 3D game may include data corresponding to a 3D object constituting the game, that is, mesh data, texture data, and the like. Based on this, the game apparatus based on the game data can extract data corresponding to the 3D object, that is, mesh data, texture data, etc. from the game data (S501).

The 3D object can be generated and displayed by rendering data corresponding to the 3D object, that is, mesh data, texture data, and the like. Furthermore, depending on the environment in which the 3D object is placed, the 3D object may be represented and configured differently. For example, the 3D object may be configured in various ways according to the setting of the direction of shooting the 3D object, the color of the 3D object, the lighting environment around the 3D object, and the like. Basically, the data corresponding to the 3D object may include information representing a basic environment for expressing the 3D object (hereinafter referred to as 'basic environment information').

Based on this, the learning device based on the game data can basically generate an image (hereinafter referred to as a 'basic image') of the 3D object based on the basic environment information.

In addition, a game data-based learning apparatus can provide an environment in which a 3D object can be modified in various environments. For example, the game device based on the game data may provide a user interface that can receive environment information from the user, which is used in constructing the 3D object. Accordingly, the learning apparatus based on the game data can confirm the environment information input by the user through the user interface, and construct a 3D object reflecting the confirmed environment information. As another example, the game data-based learning apparatus can automatically and randomly generate the environment information used in the construction of the 3D object, and can construct a 3D object reflecting the generated environment information.

Hereinafter, the learning apparatus based on the game data defines deformation environment information in order to distinguish the environment information constructed by variously modifying the environment where the 3D object is placed, from the above-described basic environment information. Also, the image of the 3D object based on the deformed environment information is defined as the deformed image.

In such an environment, the learning device based on the game data can generate a basic image, a transformed image, and the like obtained by capturing the 3D object. In particular, the environment information used in constructing the 3D object, that is, basic environment information, (S502).

Meanwhile, the image information provided in step S502 may include basic image information and deformation environment information corresponding to the basic image and the deformation image generated based on the game data. In particular, since the deformed image and the deformed environment information are information transformed from the basic image and the basic environment information, they may have predetermined characteristics corresponding to the basic image and the basic environment information. In consideration of this, the learning apparatus based on the game data can confirm the characteristic relation between the basic image and the basic environment information, the deformation image and the modified environment information, and build a learning model for the characteristic relation through learning.

More specifically, in step S503, the learning device based on the game data extracts the features of the basic image 301 and the transformed image 302, and outputs the feature maps, for example, the basic feature map 311 and the transformed feature map 312 Can be configured. In addition, a plurality of modified images may be generated, and the game device based on the game data may configure a modified feature map for each of the plurality of modified images.

In step S504, the learning device based on the game data can learn the relation between the basic feature map 311 and the deformation feature map 312 through the deep neural network to construct a learning model based on the game data. At this time, the learning apparatus based on the game data can use the above-described basic environment information and the modified environment information for learning. Furthermore, the structure of the neural network can be composed of Auto-Associative Neural Network, Restricted Boltzman Machine, and Fully Connected Network.

In step S505, the learning device based on the game data can perform a re-learning on the learning model based on the game data established in step S504, and update the neural network included therein.

Specifically, the learning device based on the game data can construct a learning model corresponding to the basic feature map 311 and the modified feature map 312 extracted from the basic image 301 and the transformed image 302, A virtual feature map can be constructed using a model. For example, when an image to be used as actual learning data to be used in an application field (hereinafter referred to as an 'learning target image') 401 is inputted, the learning device based on the game data is learned A feature map (hereinafter, referred to as a 'learning feature map') 411 can be constructed by extracting features of the target image 401.

Then, the learning device based on the game data can construct a virtual feature map 412 corresponding to the learning subject characteristic map 411 using the pre-built learning model. At this time, the virtual feature map 412 has the same semantic label as the learning feature map 411 because the learning model is applied to the learning feature map 411 to increase data.

The learning apparatus based on the game data performs fine tuning of the learning model based on the game data using the virtual characteristic map 412 with the data enhancement and the learning target characteristic map 411 .

6 is a block diagram illustrating a computing system for executing a game data based learning method and apparatus in accordance with an embodiment of the present disclosure.

6, a computing system 1000 includes at least one processor 1100, a memory 1300, a user interface input device 1400, a user interface output device 1500, (1600), and a network interface (1700).

The processor 1100 may be a central processing unit (CPU) or a memory device 1300 and / or a semiconductor device that performs processing for instructions stored in the storage 1600. Memory 1300 and storage 1600 may include various types of volatile or non-volatile storage media. For example, the memory 1300 may include a ROM (Read Only Memory) and a RAM (Random Access Memory).

Thus, the steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by processor 1100, or in a combination of the two. The software module may reside in a storage medium (i.e., memory 1300 and / or storage 1600) such as a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, You may. An exemplary storage medium is coupled to the processor 1100, which can read information from, and write information to, the storage medium. Alternatively, the storage medium may be integral to the processor 1100. [ The processor and the storage medium may reside within an application specific integrated circuit (ASIC). The ASIC may reside within the user terminal. Alternatively, the processor and the storage medium may reside as discrete components in a user terminal.

Although the exemplary methods of this disclosure are represented by a series of acts for clarity of explanation, they are not intended to limit the order in which the steps are performed, and if necessary, each step may be performed simultaneously or in a different order. In order to implement the method according to the present disclosure, the illustrative steps may additionally include other steps, include the remaining steps except for some steps, or may include additional steps other than some steps.

The various embodiments of the disclosure are not intended to be all-inclusive and are intended to illustrate representative aspects of the disclosure, and the features described in the various embodiments may be applied independently or in a combination of two or more.

In addition, various embodiments of the present disclosure may be implemented by hardware, firmware, software, or a combination thereof. In the case of hardware implementation, one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays A general processor, a controller, a microcontroller, a microprocessor, and the like.

The scope of the present disclosure is to be accorded the broadest interpretation as understanding of the principles of the invention, as well as software or machine-executable instructions (e.g., operating system, applications, firmware, Instructions, and the like are stored and are non-transitory computer-readable medium executable on the device or computer.

Claims (1)

Game data for generating image information including environmental information indicating an environment in which the 3D object is configured and an image obtained by capturing the 3D object, Based image information generating unit,
And a learning processing unit for building a learning model based on game data using the image information provided by the image information generating unit.

Images