WO2021194105A1

WO2021194105A1 - Expert simulation model training method, and device for training

Info

Publication number: WO2021194105A1
Application number: PCT/KR2021/002342
Authority: WO
Inventors: 최정렬
Original assignee: 주식회사 에스아이에이
Priority date: 2020-03-24
Filing date: 2021-02-25
Publication date: 2021-09-30
Also published as: KR20210119208A; KR102364881B1

Abstract

Disclosed is a computer program stored in a computer-readable storage medium, for resolving a previously stated problem. The computer program comprises commands for performing steps for data processing when executed by one or more processors of a computer device, and the steps may comprise: recognizing a training data set that includes label information and one or more pieces of metadata for each piece of training data, the metadata being the information which serves as the basis for allowing the label information to be assigned to the training data; and training an expert simulation model on the basis of the training data set.

Description

Expert simulation model learning method and apparatus for learning the same

The present invention relates to a method for learning a simulated expert model using a computing device, and more particularly, to performing pseudo-labeling using the simulated expert model and generating a training data set.

Conventionally, identification of an object included in image data, etc. has been made using an object identification model based on a neural network, in particular, a Convolutional Neural Network (CNN). The primary and main condition for generating a good model through training is a good quality data set. Conventionally, in general, labeling of training data for generating a training data set has been performed by a human operator. These human workers have received short training from experts or have been based on existing common-sense identification knowledge without additional training.

Since such a labeling operation provides only the correspondence between the object and the label, it is not possible to include a high level of knowledge related to labeling in the training data set. Therefore, there is a demand to create a high-quality training data set by embedding the human expert's knowledge related to object identification in the training data.

Prior art documents of the present invention are as follows.

(Patent Document 1) Republic of Korea Patent No. 10-2030027

(Patent Document 2) Japanese Patent No. 6370749

(Patent Document 3) Korean Patent Publication No. 2017-0048304

The present disclosure has been devised in response to the above-described background technology, and an object of the present disclosure is to provide a method for training an expert simulation model.

A computer program stored in a computer-readable storage medium is disclosed according to an embodiment of the present disclosure for solving the problems as described above. The computer program includes instructions that, when executed on one or more processors of a computer device, cause the following steps to be performed for data processing, the steps comprising: a learning comprising one or more metadata and label information for each of the training data recognizing a data set, wherein the metadata is information on which the label information is based on which the training data is assigned; and training an expert simulation model based on the training data set. may include.

Also, when the learning data includes an identifiable object, the metadata may include information for defining an object region of the identifiable object.

Also, the information for defining the region of the identifiable object may include shape information of the object region and parameter value information according to the shape information of the object region.

In addition, the metadata may include labeling behavior information recognized in the labeling process.

Also, the labeling behavior information may include at least one of label correction information, object region movement information, object region shape change information, and object region rotation information.

In addition, each of the one or more metadata may further include time information.

In addition, the expert simulation model may be trained to infer at least one of one or more metadata about an identifiable object or label information of the identifiable object.

In addition, the computer program, using the trained expert simulation model, performing pseudo labeling (pseudo labeling) from the image data; may further include.

In addition, the performing of the pseudo-labeling may include: extracting one or more identifiable objects from the image data; and inferring one or more metadata and the label information for each of the one or more identifiable objects included in the image data. may include.

In addition, each of the one or more metadata for each of the identifiable objects inferred by the expert simulation model includes labeling behavior information and time information inferred in the pseudo labeling process, and labeling behavior information inferred in the pseudo labeling process and the time information may be generated based on one or more metadata about the identifiable object collected up to a previous time.

The computer program may further include: receiving feedback information on the pseudo-labeling result; and re-learning the expert simulation model based on the feedback information. may further include.

An expert imitation computing device according to an embodiment of the present disclosure for solving the problems as described above is disclosed. The computing device may include: a memory; and a processor; including,

The processor recognizes a training data set including metadata and label information for each of the training data, the metadata being information on which the label information is based on the training data, the training data An expert simulation model can be trained based on the set.

Disclosed is a computer-readable recording medium storing a data structure according to an embodiment of the present disclosure for solving the problems as described above. The computer-readable recording medium includes: A computer-readable recording medium storing a data structure corresponding to a parameter of a neural network that is at least partially updated in a learning process, wherein the operation of the neural network is based at least in part on the parameter, The learning process includes: recognizing a training data set including metadata and label information for each training data, wherein the metadata is information on which the label information is a basis for giving the training data; and training an expert simulation model based on the training data set. may include.

A method for training an expert simulation model according to an embodiment of the present disclosure for solving the above-described problem is disclosed. In the method for training an expert simulation model performed in one or more processors of a computing device, the step of recognizing a training data set including metadata and label information for each training data - The metadata is the label information information serving as a basis for being given to the learning data; and training an expert simulation model based on the training data set. may include.

The present disclosure may provide a method for training an expert simulation model.

1 is a block diagram illustrating an exemplary computing device for performing an expert simulation model training method according to the present disclosure.

2 is a schematic diagram illustrating a network function for performing an expert simulation model training method according to an embodiment of the present disclosure.

3 is a block diagram illustrating an example of learning data according to the present disclosure.

4 is a block diagram illustrating an example of object region definition information according to the present disclosure.

5 illustrates an example of a recursive neural network, which is a form of a network function according to the present disclosure.

6 is a flowchart illustrating a process in which the processor according to the present disclosure trains the expert simulation model.

7 is a flowchart illustrating an example in which the expert simulation model according to the present disclosure performs pseudo labeling.

8 is a simplified, general schematic diagram illustrating an exemplary computing environment in which embodiments of the present disclosure may be implemented.

Various embodiments are now described with reference to the drawings. In this specification, various descriptions are presented to provide an understanding of the present disclosure. However, it is apparent that these embodiments may be practiced without these specific descriptions.

The terms “component,” “module,” “system,” and the like, as used herein, refer to a computer-related entity, hardware, firmware, software, a combination of software and hardware, or execution of software. For example, a component can be, but is not limited to being, a process running on a processor, a processor, an object, a thread of execution, a program, and/or a computer. For example, both an application running on a computing device and the computing device may be a component. One or more components may reside within a processor and/or thread of execution. A component may be localized within one computer. A component may be distributed between two or more computers. In addition, these components can execute from various computer readable media having various data structures stored therein. Components may communicate via a network such as the Internet with another system, for example via a signal having one or more data packets (eg, data and/or signals from one component interacting with another component in a local system, distributed system, etc.) may communicate via local and/or remote processes depending on the data being transmitted).

In addition, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or.” That is, unless otherwise specified or clear from context, "X employs A or B" is intended to mean one of the natural implicit substitutions. That is, X employs A; X employs B; or when X employs both A and B, "X employs A or B" may apply to either of these cases. It should also be understood that the term “and/or” as used herein refers to and includes all possible combinations of one or more of the listed related items.

Also, the terms "comprises" and/or "comprising" should be understood to mean that the feature and/or element in question is present. However, it should be understood that the terms "comprises" and/or "comprising" do not exclude the presence or addition of one or more other features, elements and/or groups thereof. Also, unless otherwise specified or unless the context is clear as to designating a singular form, the singular in the specification and claims should generally be construed to mean "one or more."

In addition, the term “at least one of A or B” should be interpreted as meaning “includes only A”, “includes only B”, and “in the case of a combination of A and B”.

Those skilled in the art will further appreciate that the various illustrative logical blocks, configurations, modules, circuits, means, logics, and algorithm steps described in connection with the embodiments disclosed herein may be implemented in electronic hardware, computer software, or combinations of both. It should be recognized that they can be implemented with To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, configurations, means, logics, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application. However, such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

Descriptions of the presented embodiments are provided to enable those skilled in the art to use or practice the present invention. Various modifications to these embodiments will be apparent to those skilled in the art of the present disclosure. The generic principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Thus, the present invention is not limited to the embodiments presented herein. This invention is to be construed in its widest scope consistent with the principles and novel features presented herein.

1 is a block diagram of a computing device for performing a method for generating an expert simulation model according to an embodiment of the present disclosure.

The expert simulation model according to the present disclosure may be a neural network that infers label information and metadata for an identifiable object included in image data by learning the labeling process of a human expert. The expert simulation model may be a model that receives image data including one or more identifiable objects and infers label information and metadata for the identifiable object 410 . Accordingly, the training data 400 may include time information and may be configured in the form of time series data.

The configuration of the computing device 100 shown in FIG. 1 is only a simplified example. In an embodiment of the present disclosure, the computing device 100 may include other components for performing the computing environment of the computing device 100 , and only some of the disclosed components may configure the computing device 100 .

The computing device 100 may include a processor 110 and a memory 120 .

The processor 110 may include one or more cores, and a central processing unit (CPU) of a computing device, a general purpose graphics processing unit (GPGPU), and a tensor processing unit (TPU). unit) and the like, and may include a processor for data analysis and deep learning. The processor 110 may read a computer program stored in the memory 130 and perform data processing for machine learning according to an embodiment of the present disclosure. According to an embodiment of the present disclosure, the processor 110 may perform an operation for learning the neural network. The processor 110 for learning of the neural network, such as processing input data for learning in deep learning (DL), extracting features from the input data, calculating an error, updating the weight of the neural network using backpropagation calculations can be performed. At least one of a CPU, a GPGPU, and a TPU of the processor 110 may process learning of a network function. For example, the CPU and the GPGPU can process learning of a network function and data classification using the network function together. In addition, in an embodiment of the present disclosure, learning of a network function and data classification using the network function may be processed by using the processors of a plurality of computing devices together. In addition, the computer program executed in the computing device according to an embodiment of the present disclosure may be a CPU, GPGPU, or TPU executable program.

According to an embodiment of the present disclosure, the memory 120 may store any type of information generated or determined by the processor 110 and any type of information received by the network unit.

According to an embodiment of the present disclosure, the memory 120 is a flash memory type, a hard disk type, a multimedia card micro type, or a card type memory (eg, SD or XD memory, etc.), Random Access Memory (RAM), Static Random Access Memory (SRAM), Read-Only Memory (ROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Programmable Read (PROM) -Only Memory), a magnetic memory, a magnetic disk, and an optical disk may include at least one type of storage medium.

2 is a schematic diagram illustrating a network function for performing a method for generating an expert simulation model according to an embodiment of the present disclosure.

Throughout this specification, computational model, neural network, network function, and neural network may be used interchangeably. A neural network may be composed of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network is configured by including at least one or more nodes. Nodes (or neurons) constituting the neural networks may be interconnected by one or more links.

In the neural network, one or more nodes connected through a link may relatively form a relationship between an input node and an output node. The concepts of an input node and an output node are relative, and any node in an output node relationship with respect to one node may be in an input node relationship in a relationship with another node, and vice versa. As described above, an input node-to-output node relationship may be created around a link. One or more output nodes may be connected to one input node through a link, and vice versa.

In the relationship between the input node and the output node connected through one link, the value of the data of the output node may be determined based on data input to the input node. Here, a link interconnecting the input node and the output node may have a weight. The weight may be variable, and may be changed by the user or algorithm in order for the neural network to perform a desired function. For example, when one or more input nodes are interconnected to one output node by respective links, the output node sets values input to input nodes connected to the output node and links corresponding to the respective input nodes. An output node value may be determined based on the weight.

As described above, in a neural network, one or more nodes are interconnected through one or more links to form an input node and an output node relationship within the neural network. The characteristics of the neural network may be determined according to the number of nodes and links in the neural network, the correlation between the nodes and the links, and the value of a weight assigned to each of the links. For example, when the same number of nodes and links exist and there are two neural networks having different weight values of the links, the two neural networks may be recognized as different from each other.

A neural network may consist of a set of one or more nodes. A subset of nodes constituting the neural network may constitute a layer. Some of the nodes constituting the neural network may configure one layer based on distances from the initial input node. For example, a set of nodes having a distance n from the initial input node may constitute n layers. The distance from the initial input node may be defined by the minimum number of links that must be traversed to reach the corresponding node from the initial input node. However, the definition of such a layer is arbitrary for description, and the order of the layer in the neural network may be defined in a different way from the above. For example, a layer of nodes may be defined by a distance from the final output node.

The initial input node may refer to one or more nodes to which data is directly input without going through a link in a relationship with other nodes among nodes in the neural network. Alternatively, in a relationship between nodes based on a link in a neural network, it may mean nodes that do not have other input nodes connected by a link. Similarly, the final output node may refer to one or more nodes that do not have an output node in relation to other nodes among nodes in the neural network. In addition, the hidden node may mean nodes constituting the neural network other than the first input node and the last output node.

The neural network according to an embodiment of the present disclosure may be a neural network in which the number of nodes in the input layer may be the same as the number of nodes in the output layer, and the number of nodes decreases and then increases again as progresses from the input layer to the hidden layer. can In addition, the neural network according to another embodiment of the present disclosure may be a neural network in which the number of nodes in the input layer may be less than the number of nodes in the output layer, and the number of nodes decreases as the number of nodes progresses from the input layer to the hidden layer. have. In addition, the neural network according to another embodiment of the present disclosure may be a neural network in which the number of nodes in the input layer may be greater than the number of nodes in the output layer, and the number of nodes increases as the input layer progresses to the hidden layer. can The neural network according to another embodiment of the present disclosure may be a neural network in a combined form of the aforementioned neural networks.

A deep neural network (DNN) may refer to a neural network including a plurality of hidden layers in addition to an input layer and an output layer. Deep neural networks can be used to identify the latent structures of data. In other words, it can identify the potential structure of photos, texts, videos, voices, and music (for example, what objects are in the photos, what the text and emotions are, what the texts and emotions are, etc.) . Deep neural networks include convolutional neural networks (CNNs), recurrent neural networks (RNNs), auto encoders, generative adversarial networks (GANs), and restricted boltzmann machines (RBMs). machine), a deep belief network (DBN), a Q network, a U network, a Siamese network, a Generative Adversarial Network (GAN), and the like. The description of the deep neural network described above is only an example, and the present disclosure is not limited thereto.

In an embodiment of the present disclosure, the network function may include an autoencoder. The auto-encoder may be a kind of artificial neural network for outputting output data similar to input data. The auto encoder may include at least one hidden layer, and an odd number of hidden layers may be disposed between the input/output layers. The number of nodes in each layer may be reduced from the number of nodes of the input layer to an intermediate layer called the bottleneck layer (encoding), and then expanded symmetrically with reduction from the bottleneck layer to the output layer (symmetrically with the input layer). The auto-encoder can perform non-linear dimensionality reduction. The number of input layers and output layers may correspond to a dimension after preprocessing the input data. In the auto-encoder structure, the number of nodes of the hidden layer included in the encoder may have a structure that decreases as the distance from the input layer increases. If the number of nodes in the bottleneck layer (the layer with the fewest nodes between the encoder and the decoder) is too small, a sufficient amount of information may not be conveyed, so a certain number or more (e.g., more than half of the input layer, etc.) ) may be maintained.

The neural network may be trained using at least one of supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning. Learning of the neural network may be a process of applying knowledge for the neural network to perform a specific operation to the neural network.

A neural network can be trained in a way that minimizes output errors. In the training of a neural network, iteratively inputs the training data to the neural network, calculates the output of the neural network and the target error for the training data, and calculates the error of the neural network from the output layer of the neural network to the input layer in the direction to reduce the error. It is a process of updating the weight of each node in the neural network by backpropagation in the direction. In the case of teacher learning, learning data in which the correct answer is labeled in each learning data is used (ie, labeled learning data), and in the case of comparative learning, the correct answer may not be labeled in each learning data. That is, for example, the learning data in the case of teacher learning regarding data classification may be data in which categories are labeled in each of the learning data. Labeled training data is input to the neural network, and an error can be calculated by comparing the output (category) of the neural network with the label of the training data. As another example, in the case of comparison learning about data classification, an error may be calculated by comparing the input training data with the neural network output. The calculated error is back propagated in the reverse direction (ie, from the output layer to the input layer) in the neural network, and the connection weight of each node of each layer of the neural network may be updated according to the back propagation. A change amount of the connection weight of each node to be updated may be determined according to a learning rate. The computation of the neural network on the input data and the backpropagation of errors can constitute a learning cycle (epoch). The learning rate may be applied differently according to the number of repetitions of the learning cycle of the neural network. For example, in the early stage of training of a neural network, a high learning rate can be used to enable the neural network to quickly acquire a certain level of performance, thereby increasing efficiency, and using a low learning rate at the end of learning can increase accuracy.

In the training of neural networks, in general, the training data may be a subset of real data (that is, data to be processed using the trained neural network), and thus the error on the training data is reduced, but the error on the real data is reduced. There may be increasing learning cycles. Overfitting is a phenomenon in which errors on actual data increase by over-learning on training data as described above. For example, a phenomenon in which a neural network that has learned a cat by showing a yellow cat does not recognize that it is a cat when it sees a cat other than yellow may be a type of overfitting. Overfitting can act as a cause of increasing errors in machine learning algorithms. In order to prevent such overfitting, various optimization methods can be used. In order to prevent overfitting, methods such as increasing the training data, regularization, and dropout that deactivate some of the nodes of the network in the process of learning, and the use of a batch normalization layer are applied. can

A computer-readable medium storing a data structure is disclosed according to an embodiment of the present disclosure.

The data structure may refer to the organization, management, and storage of data that enables efficient access and modification of data. A data structure may refer to an organization of data to solve a specific problem (eg, data retrieval, data storage, and data modification in the shortest time). A data structure may be defined as a physical or logical relationship between data elements designed to support a particular data processing function. The logical relationship between data elements may include a connection relationship between user-defined data elements. Physical relationships between data elements may include actual relationships between data elements physically stored on a computer-readable storage medium (eg, persistent storage). A data structure may specifically include a set of data, relationships between data, and functions or instructions applicable to data. Through an effectively designed data structure, a computing device can perform an operation while using the resources of the computing device to a minimum. Specifically, the computing device may increase the efficiency of operations, reads, insertions, deletions, comparisons, exchanges, and retrievals through effectively designed data structures.

A data structure may be classified into a linear data structure and a non-linear data structure according to the type of the data structure. The linear data structure may be a structure in which only one piece of data is connected after one piece of data. The linear data structure may include a list, a stack, a queue, and a deck. A list may mean a set of data in which an order exists internally. The list may include a linked list. The linked list may be a data structure in which data is linked in such a way that each data is linked in a line with a pointer. In a linked list, a pointer may contain information about a link with the next or previous data. A linked list may be expressed as a single linked list, a doubly linked list, or a circularly linked list according to a shape. A stack can be a data enumeration structure with limited access to data. A stack can be a linear data structure in which data can be processed (eg, inserted or deleted) at only one end of the data structure. The data stored in the stack may be a data structure LIFO-Last in First Out. A queue is a data listing structure that allows limited access to data. Unlike a stack, the queue may be a data structure that comes out later (FIFO-First in First Out) as data stored later. A deck can be a data structure that can process data at either end of the data structure.

The nonlinear data structure may be a structure in which a plurality of data is connected after one data. The nonlinear data structure may include a graph data structure. A graph data structure may be defined as a vertex and an edge, and the edge may include a line connecting two different vertices. A graph data structure may include a tree data structure. The tree data structure may be a data structure in which one path connects two different vertices among a plurality of vertices included in the tree. That is, it may be a data structure that does not form a loop in the graph data structure.

Throughout this specification, computational model, neural network, network function, and neural network may be used interchangeably. Hereinafter, the neural network is unified and described. The data structure may include a neural network. And the data structure including the neural network may be stored in a computer-readable medium. Data structures, including neural networks, also include preprocessed data for processing by the neural network, data input to the neural network, weights of the neural network, hyperparameters of the neural network, data obtained from the neural network, activation functions associated with each node or layer of the neural network, and the neural network. It may include a loss function for learning of . A data structure comprising a neural network may include any of the components disclosed above. That is, the data structure including the neural network includes preprocessed data for processing by the neural network, data input to the neural network, weights of the neural network, hyperparameters of the neural network, data obtained from the neural network, activation functions associated with each node or layer of the neural network, and the neural network It may be configured to include all or any combination thereof, such as a loss function for learning of . In addition to the above-described configurations, a data structure including a neural network may include any other information that determines a characteristic of a neural network. In addition, the data structure may include all types of data used or generated in the operation process of the neural network, and is not limited to the above. Computer-readable media may include computer-readable recording media and/or computer-readable transmission media. A neural network may be composed of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network is configured by including at least one or more nodes.

The data structure may include data input to the neural network. A data structure including data input to the neural network may be stored in a computer-readable medium. The data input to the neural network may include learning data input in a neural network learning process and/or input data input to the neural network in which learning is completed. Data input to the neural network may include pre-processing data and/or pre-processing target data. The preprocessing may include a data processing process for inputting data into the neural network. Accordingly, the data structure may include data to be pre-processed and data generated by pre-processing. The above-described data structure is merely an example, and the present disclosure is not limited thereto.

The data structure may include the weights of the neural network. (In this specification, a weight and a parameter may be used interchangeably.) And a data structure including a weight of a neural network may be stored in a computer-readable medium. The neural network may include a plurality of weights. The weight may be variable, and may be changed by the user or algorithm in order for the neural network to perform a desired function. For example, when one or more input nodes are interconnected to one output node by respective links, the output node sets values input to input nodes connected to the output node and links corresponding to the respective input nodes. A data value output from the output node may be determined based on the weight. The above-described data structure is merely an example, and the present disclosure is not limited thereto.

By way of example and not limitation, the weight may include a weight variable in a neural network learning process and/or a weight in which neural network learning is completed. The variable weight in the neural network learning process may include a weight at the start of the learning cycle and/or a variable weight during the learning cycle. The weight for which neural network learning is completed may include a weight for which a learning cycle is completed. Accordingly, the data structure including the weight of the neural network may include a data structure including the weight variable in the neural network learning process and/or the weight in which the neural network learning is completed. Therefore, it is assumed that the above-described weights and/or combinations of weights are included in the data structure including the weights of the neural network. The above-described data structure is merely an example, and the present disclosure is not limited thereto.

The data structure including the weights of the neural network may be stored in a computer-readable storage medium (eg, memory, hard disk) after being serialized. Serialization can be the process of converting a data structure into a form that can be reconstructed and used later by storing it on the same or a different computing device. The computing device may serialize the data structure to send and receive data over the network. A data structure including weights of the serialized neural network may be reconstructed in the same computing device or in another computing device through deserialization. The data structure including the weights of the neural network is not limited to serialization. Furthermore, the data structure including the weights of the neural network is a data structure to increase the efficiency of computation while using the resources of the computing device to a minimum (e.g., B-Tree, Trie, m-way search tree, AVL tree, Red-Black Tree). The foregoing is merely an example, and the present disclosure is not limited thereto.

The data structure may include hyper-parameters of the neural network. In addition, the data structure including the hyperparameters of the neural network may be stored in a computer-readable medium. The hyperparameter may be a variable variable by a user. Hyperparameters are, for example, learning rate, cost function, number of iterations of the learning cycle, weight initialization (e.g., setting the range of weight values subject to weight initialization), Hidden Unit The number (eg, the number of hidden layers, the number of nodes of the hidden layer) may be included. The above-described data structure is merely an example, and the present disclosure is not limited thereto.

The training data 400 according to the present disclosure may include image data including the identifiable object 410 , label information, and metadata.

The training data 400 may be data input to the neural network during training of the expert simulation model according to the present disclosure. The training data 400 may be previously generated and stored in the memory 120 irrespective of the expert simulation model according to the present disclosure.

Alternatively, the training data 400 may be generated based on metadata and label information for the identifiable object 410 inferred by the expert simulation model according to the present disclosure. For example, the processor 110 may reconstruct the learning data based on feedback information received with respect to the inferred metadata and label information. Details on this will be described later with reference to FIG. 6 .

Exemplarily, the training data 400 includes an identifiable object 410, label information 420 at a previous time, label information 430 at a current time, labeling behavior information 440 at a current time, and a current time. may include object region definition information 450 in .

In the present disclosure, the identifiable object 410 may mean a single object recognized separately from the background in the image data. For example, the identifiable object 410 may be an airplane, a tank, a ship, a vehicle, and an oil storage tank. The description of the identifiable object described above is merely an example, and the present disclosure is not limited thereto.

In the present disclosure, the

label information

420 and 430 may correspond to “none”.

The training data according to the present disclosure is characterized in that it accompanies not only label information but also labeling-related metadata. That is, metadata information generated before label information is given to the identifiable object 410 may be generated in relation to the identifiable object 410 .

To this end, the processor 110 may recognize a time point at which the label information is initially assigned, and collect metadata for a predefined time period from the time point.

The predefined time interval time may include both a time interval before and a time interval after the time point to which the first label information is given.

Alternatively, the processor 110 may recognize a point in time at which the cursor is positioned within a predetermined distance from the identifiable object 410 to which the label is attached during the labeling process. The processor 110 may configure the learning data 400 using metadata from a point in time when the cursor is positioned within a predetermined distance from an identifiable object.

The processor 110 may configure the learning data 400 by setting the label information to “none” for the collected metadata information.

Since this is only an example of configuring the training data 400 , the method of configuring the training data 400 is not limited thereto.

Meta data according to the present disclosure may be information serving as a basis for providing the label information to the learning data. The metadata may be information generated in a process of labeling an identifiable object. One or more metadata may be provided to each of the identifiable objects 410 . When one or more meta data per one identifiable object 410 exists, each of the meta data may represent a labeling process at different viewpoints.

For example, it is assumed that three pieces of metadata exist for the identifiable object 410 . In this case, the first metadata may include labeling behavior information and object region definition information at the first time point, and the second metadata may include labeling behavior information and object region definition information at the second time point.

In the present disclosure, the object area may mean an area occupied by the identifiable object 410 on image data. These object regions may have various geometric shapes. For example, the object region may be a circle, an ellipse, a triangle, a rectangle, a polygon, or the like.

The object region may be input through an input device and recognized by the processor 110 . Alternatively, it may be recognized using an object detection and localization technology in a computer vision area. The above-described object recognition technology is merely an example, and the present disclosure is not limited thereto.

When the learning data includes an identifiable object, the metadata may include object region definition information of the identifiable object. The object area may be defined by the object area definition information 450 . The object region definition information may include information for describing the object region. For example, when the object region is a circle, the object region definition information may include center pixel coordinates of the circle and radius length information of the circle. When the object area is a rectangle, the object area definition information may include coordinates of four vertices of the rectangle.

The object region definition information may be different according to the type of the object region. Accordingly, the above-described form of the object region definition information is exemplary, and the object region definition information is not limited thereto.

The metadata may include labeling behavior information on the identifiable object 410 . In the present disclosure, labeling action information may be information representing actions performed in a labeling process by a labeling task performing subject. That is, the labeling action information may include an action of an identifiable object regarding an object region, an action of adjusting the label information of an identifiable object, and the like.

Specifically, the labeling behavior information may include at least one of label correction information, object region movement information, object region shape change information, and object region rotation information. In addition, the labeling behavior information includes information on the action of zooming in on the image, information on the action of zooming out, and information on the action of adjusting the contrast for understanding the contrast such as adjusting the color to black and white. can do. For example, the labeling action information may include information about actions performed by designating an object area for an object identifiable by a user, setting a label, and performing the labeling action information. For example, the labeling behavior information allows the user to designate a first object area for a first identifiable object, assign a first label, modify the first object area to set a second object area for a second identifiable object, and When a second label is assigned to this, information related to correction and change of the second object area and the second label in the first object area and the first label may be included.

The above description is only an example of labeling behavior information. Accordingly, the labeling behavior information should not be limited to the above-described example.

The label correction information may mean whether a label correction has been made on the identifiable object 410 . For example, if the label information is modified, the label information may be 1, otherwise, the label modification information may be 0.

Exemplarily, when the class of label information at the previous time point is different from the class of the label information at the current time point, the label modification information may include information related to class modification of the label. The information related to class modification may include, for example, whether a class modification has occurred, or specifically how the class has been modified.

For example, the label modification information may include information on how the class is specifically modified by comparing the label information 420 at the previous time point with the label information 430 at the current time point.

For example, the memory 120 may store a plurality of candidate label information and label change type information for the identifiable object 410 . Specifically, the label change type information may include information that, when the label information for the identifiable object 410 is changed from "fighter" to "airliner", this change of label information corresponds to "3". . This is only an example of the label change information, and the expression method of the label change information is not limited thereto.

In the present disclosure, the object region movement information may mean whether the object region defined for an identifiable object has been moved. Exemplarily, the object region movement information may include information related to the movement performed when a movement to an area established as an identifiable object occurs in the labeling process. For example, the object region movement information may be 1 when there is a movement of the object region, and 0 otherwise.

As another example, the object region movement information may mean a change in the position of the reference point of the object region. The case where the shape of the object area is the cause will be described as an example. When the shape of the object region is the cause, the reference point of the object region may be the coordinates of the center of the circle. In this case, the object region movement information may be expressed as a change in position between the center coordinates of the circle at the previous time point and the center coordinates of the circle at the current view point. If the center coordinates of the circle at the first time point are (3,4) and the center coordinates of the circle at the second time point are (1,2), the object area movement information may be (-2, -2). have.

The processor 110 according to the present disclosure may generate shape change information of an object area and include it in metadata. The shape change information of the object area may represent whether the shape of the object area is changed. Furthermore, the shape change information of the object area may represent the type of shape change of the object area.

For example, when the shape change of the object area is not performed, the processor 110 may determine the shape change information of the object area as 0. For example, when the user sets the object area in the shape of a rectangular box with respect to the airplane object and then changes the object area to the shape of an airplane, the shape change information of the object area compares the shape change information with the change information before and after the change may include information about the , information related to detailed operations of the performed change, and the like.

The processor 110 may correspond the type of change in the shape of the object area to a specific data value according to a preset rule. For example, when the shape of the object region is changed from a rectangle to a circle, the shape change information of the object region may be determined to be 1. When the shape of the object region is changed from a circle to an ellipse, shape change information of the object region may be determined as 2 .

The object region rotation information according to the present disclosure may represent whether the object region is rotated and the degree of rotation of the object region.

For example, when rotation of the object region is not performed, the object region rotation information may be 0.

When the object region is rotated by 90 degrees in the clockwise direction, the processor 110 may determine the object region rotation information to be +90. Conversely, when the rotation of the object region is 90 degrees counterclockwise, the processor 110 may determine the object region rotation information to be -90. Also, for example, the rotation information of the object region is information related to the modification performed by the user when the user sets the second object region in the second identifiable object by rotating the first object region set in the first identifiable object. may include.

At least one metadata including at least one of labeling behavior information and object region definition information may exist for an identifiable object. In this case, one piece of metadata may correspond to a point in the labeling process. To express this, each of the metadata may include time information. Through this, it is possible to compose a training data set by identifying changes in metadata and label information in the labeling process over time.

The learning data 400 disclosed in FIG. 3 will be described as an example.

The training data 400 may include an identifiable object 410 . In the example shown through FIG. 3 , the identifiable object 410 is an airliner.

The training data 400 shown at the far left is the training data at the first time point, that is, the time t=t_0, and in this case, time information is 0.

As described above, in the present disclosure, the

label information

420 and 430 may correspond to “none”.

To this end, the processor 110 may recognize a time point at which the label information is initially assigned, and collect metadata from the time point before a predefined time interval.

Based on this, the label information 430 at t=t_0 or the label information 420 at t=t_1 may still correspond to “none”. Furthermore, the

label information

420 and 430 may correspond to “none” until the time when the label information 430 is given by the expert or the processor 110 configuring the learning data 400 .

Based on this, it is possible to collect metadata from a point in time when the subject constituting the learning data 400 recognizes the identifiable object 410 as an identification target. Accordingly, the amount of information contained in the learning data 400 may be further enriched.

However, hereinafter, for convenience of explanation, it is assumed that label information is first assigned from t=t_0.

The training data 400 may include label information 430 at the current time point. In the example shown in FIG. 3 , the label information (label information at t=t_0) 430 at the current time point may be a “fighter”.

Furthermore, at the next point in time, that is, when the time information is 1, the label information 420 of the previous time may be a "fighter", and in this case, the label information at the current time (label information at t = t_1) 430 may be a "passenger plane".

Also, the metadata may include labeling behavior information recognized in the labeling process.

In the example of time t=t_0 described above, the label 420 at the previous time point may be “ship”, and the label at the current time point may be “fighter”. In this case, the label correction information at the time t=t_0 may be 1.

Reference is made to the identifiable object 410 at time points t=t_1 and t=t_2. When the information defining the object area at the time t=t_1 is four vertices of the rectangle, the central point of the four vertices may correspond to the reference point.

Object region movement information may be determined based on reference point coordinates (2,2) 452a at time t=t_1 and reference point coordinates (3,4) 452b at time t=t_2. In this case, the object region movement information at the time t=t_2 may be (3 - 2, 4 - 2), that is, (1,2).

An object area of any type may be allocated at an initial time for an identifiable object. At t=t_-1, the allocated object area is a circle, and at t=t_0 it has been changed to a rectangle. If the shape change from a circle to a rectangle corresponds to “5” according to a preset rule, the object region shape change information at t=t_1 may be “5”.

Referring to FIG. 3 , a quadrangle at t=t_1 is an initially allocated quadrangle and is not rotated. In the rightmost training data of FIG. 3 , the object area may be rotated counterclockwise. If the rotation angle is 45 degrees, the object region rotation information at t=t_2 may be “-45”.

The process of recognizing and labeling a visual object includes a process of separating the object and the background, recognizing the area occupied by the object, and identifying the shape of the area occupied by the object.

For a visually identifiable object, the amount of information available to the recognition and labeling model for training is limited in the case of providing a correspondence between the image and the label only. Also, it is very difficult to interpret the output data.

By configuring the training data as described above, it is possible to sufficiently provide the evidence data for labeling when the model is trained, and to allow the model to output the evidence data for labeling. Thereby, it is possible to provide sufficient data for training of the labeling model, and to obtain sufficient interpretation grounds for the labeling inference result of the model. In addition, by allowing the model to learn about the labeling process, errors that may occur in the labeling process, difficulty of each labeling process, etc. can be learned. Accordingly, the model may vary the degree of learning according to the difficulty of each process of labeling, and the model may have interpretability of the inference result of the model by simulating and inferring the process of labeling in the process of inferring.

Also, recognition and labeling of visual objects occurs during arbitrary time intervals. In response to this, the training data may be configured time-series. By organizing the training data in a time series, the classification model can more closely mimic the expert's labeling process. You can also look at the labeling process of the classification model, and better understand where the classification model makes a misjudgment.

The object region definition information 450 may include shape information of the object region and parameter value information according to the shape information of the object region.

As described above, the shape information 451 of the object region may be included in the learning data 400 . The shape information 451 of the object region may be included in the training data 400 and already stored in the memory 120 in order to perform the training method of the expert simulation model according to the present disclosure.

Specifically, the shape information 451 of the object region may be a polygon, a circle, or the like. The parameter information 452 according to the object region shape information 451 may be different according to the shape information 451 . For example, when the object region shape information is a polygon, the parameter information 452 may be coordinates of vertices. When the object region shape information is the cause, the parameter information 452 may be the location of the origin and the size of the radius.

The shape information 451 and parameter information 452 of the object region may be input through an input device during the labeling process.

Also, the input may be input using object detection and localization techniques in the computer vision field.

For example, in step S300 of FIG. 6 to be described later, the expert model according to the present disclosure may extract an identifiable object from image data for labeling the identifiable object 410 . When the image data includes a plurality of identifiable objects, the processor 110 may detect the plurality of objects included in the image data. Also, the processor 110 may detect an object region including each object. Based on this, the processor 110 may detect the shape information 451 of the object area including each object and the parameter information 452 according to the object area shape information.

In recognizing and labeling a visual object, it is important to obtain an object region including the most features of the object. If the shape of the object area is not suitable for the object to be identified, an excessive amount of unnecessary background may be included in the object area, or the characteristics of the object to be identified may not be sufficiently contained in the object area. Accordingly, by including the shape information of the object region in the training data and parameters for well representing the object region, an expert simulation model capable of inferring an appropriate object region may be generated.

5 shows an example of a recurrent neural network in the form of an artificial neural network according to the present disclosure.

As shown in FIG. 5 , in the present disclosure, the network function may have a form of a recurrent neural network (RNN) as well as a form of a general artificial neural network. A cyclic neural network has a characteristic that connections between units have a cyclic structure. This structure allows the state to be stored inside the neural network to model time-varying dynamic features. Unlike forward forward neural networks, recurrent neural networks can process input in the form of sequences using internal memory. Therefore, the cyclic neural network can process data having time-varying characteristics, such as handwriting recognition and speech recognition. The description of the above-mentioned data is merely an example, and the present disclosure is not limited thereto.

The input data according to the present disclosure is data input to a neural network. In particular, when the neural network is a recurrent neural network, the input data may be metadata and label information.

Output data according to the present disclosure is a result of input data being derived through a network function, and may be a value corresponding to metadata and label information derived by the network function at the current time point. For example, when the expert simulation model generating method according to the present disclosure relates to identification and labeling of an object, the output data Y is metadata and label for an identifiable object at the current time, derived by a neural network. may contain information.

Referring to FIG. 5 , when input data X is input to a cyclic neural network, output data Y is calculated as a result. Also, as shown in FIG. 5 , when the neural network takes the form of a recurrent neural network, the unit and output data Y of the recurrent neural network may affect the operation of the next unit.

For example, suppose that the output data Y represents metadata and label information at the current time point predicted based on a sequence of metadata and label information collected up to a previous time point. When meta data and label information of a previous point in time are input to the recurrent neural network A as input data X, the output data Y may represent inferred values of meta data and label information at the specific point in time.

The structure of the recurrent neural network can continuously derive metadata and label information values according to time by using the above method.

RNNs are generally suitable for modeling sequence/time series data. Accordingly, the input data X and the output data Y may be related to metadata over time. This is only an example regarding the type of sequence/time series data, and the type of sequence/time series data is not limited thereto.

Since the above description is only an example regarding the types of input data and output data, the input data and output data are not limited to the above-described examples.

The expert simulation model may be trained to infer at least one of one or more metadata about an identifiable object or label information of the identifiable object.

The expert simulation model according to the present disclosure may be a neural network that infers label information and metadata for an identifiable object included in image data by learning the labeling process of a human expert. The simulated expert model according to the present disclosure may be based on a recurrent neural network. The simulated expert model receives image data including one or more identifiable objects and infers label information and metadata for the identifiable object 410 . It can be a model that does it. Accordingly, the training data 400 may include time information and may be configured in the form of time series data.

When the training data is configured in the form of time series data, information on the labeling process of the identifiable object 410 may be included in the training data 400 .

As mentioned above, the recognition and labeling of visual objects occurs during any time interval. In response to this, the training data may be configured time-series. By organizing the training data in a time series, the classification model can more closely mimic the expert's labeling process. You can also look at the labeling process of the classification model, and better understand where the classification model makes a misjudgment.

In the present disclosure, the identifiable object 410 may mean a single object recognized separately from the background in the image data. For example, the identifiable object 410 may be an airplane, a tank, a ship, a vehicle, and an oil storage tank. The above-described identifiable object is merely an example, and the present disclosure is not limited thereto.

The inferred data Y according to the present disclosure may include metadata and label information for an identifiable object. Like the training data 400 , a plurality of inferred data Y may exist for one identifiable object.

When the processor 110 derives a plurality of inferred data for one identifiable object, an expert may determine whether the inference model has been inferred through an appropriate process. In addition, if there is an object whose identification is ambiguous, an expert can examine the labeling process and get hints or review the feasibility of pseudo-labeling.

The processor 110 may perform pseudo labeling from image data by using the trained expert simulation model.

In the present disclosure, pseudo labeling may refer to classifying one or more identifiable object classes included in image data using an expert simulation model that has been trained. A high degree of labeling automation may be possible by using pseudo-labeling of image data generated by trained expert simulation models.

In addition, the expert simulation model generated as described above can present not only simple label information, but also the process until the model determines the label in the form of metadata. Therefore, the possibility of explaining the results derived by the model is increased, so that the understanding of the model can be further improved, and furthermore, it can be easier to derive and improve the error of the model.

The processor 110 may infer only label information for an identifiable object.

Alternatively, the processor 110 extracts one or more identifiable objects from the image data, and performs pseudo labeling by inferring one or more metadata and the label information for each of the one or more identifiable objects included in the image data. can be done

The extraction of identifiable objects using the expert simulation model can be performed using object detection and localization techniques in the computer vision domain. Therefore, if necessary, the expert simulation model according to the present disclosure includes a module for extracting a plurality of identifiable objects from image data, and a module for generating metadata and label information for the object from image data including the identifiable objects may include all of them.

Each of the one or more metadata for each of the identifiable objects inferred by the expert simulation model includes labeling behavior information and time information inferred in the pseudo labeling process, and labeling behavior information and time inferred in the pseudo labeling process The information may be generated based on one or more metadata about the identifiable object collected up to a previous time.

In the present disclosure, labeling action information may be information representing actions performed in a labeling process by a labeling task performing subject. That is, the labeling action information may include an action of an identifiable object regarding an object region, an action of adjusting the label information of an identifiable object, and the like. In addition, the labeling behavior information includes information on the action of zooming in on the image, information on the action of zooming out, and information on the action of adjusting the contrast for understanding the contrast such as adjusting the color to black and white. can do.

As described above, the processor 110 may generate label information and metadata about the identifiable object based on one or more metadata about the identifiable object collected up to a previous time.

That is, when the input data (X) is input to the cyclic neural network, the output data (Y) is calculated as a result. Also, as shown in FIG. 5 , when the neural network takes the form of a recurrent neural network, the unit and output data Y of the recurrent neural network may affect the operation of the next unit.

Hereinafter, it will be described in detail with reference to FIG. 5 .

When the recurrent neural network structure as shown in FIG. 5 is included in at least a part of the expert simulation model according to the present disclosure, the output data Y1 at the first time point may be determined as the input data X1 at the second time point.

In addition, due to the nature of the cyclic artificial neural network structure, the operation processed by the neural network A at one point in time affects the structure of the neural network A at the next point in time. Therefore, the structure of the neural network (A) at a specific point in time includes information on operations made through the neural network (A) up to the previous point in time. That is, information on the input and output meta data (X, Y) through the neural network (A) is continuously transmitted.

Accordingly, the output data Yt at the t-th time may be a result obtained by sequentially inputting one or more meta data included in the output data collected up to the previous time point to the neural network A.

Since this is only an example of a method for generating meta data, the method for generating meta data is not limited to the above description.

Referring to FIG. 6 , the processor may recognize a training data set including one or more metadata and label information for each training data ( S100 ).

The training data may be data input to the neural network during training of the expert simulation model according to the present disclosure. The training data 400 may be previously generated and stored in the memory 120 irrespective of the expert simulation model according to the present disclosure.

The processor may train an expert simulation model based on the recognized training data set (S200).

Training of expert simulation models can be done based on training data already generated by human operators.

The processor may perform pseudo-labeling from the image data using the trained expert simulation model (S300).

The processor may receive feedback information on the pseudo-labeling result (S400).

In the present disclosure, the feedback information may be correction information on data inferred by the trained expert simulation model. For example, even when the inferred data includes only label information, the feedback information may include label information and metadata determined by a subject generating the feedback information.

Specifically, even when the inferred data includes only label information “fighter” for an identifiable object, the feedback information may include changes to the fighter label, object area definition information, and labeling behavior information.

The processor may retrain the expert simulation model based on the feedback information (S500).

The processor 110 according to some embodiments of the present disclosure may evaluate whether it is appropriate to reflect the expert feedback information before re-learning.

Specifically, the processor may determine whether to relearn by comparing the labeling result or the labeling policy based on the expert's feedback information and the pseudo-labeling result generated by the expert simulation model.

For example, the processor 110 may generate a feedback reflecting labeling result by reflecting the received feedback information on the generated pseudo labeling result.

The processor 110 may compare accuracy between the feedback-reflected labeling result and the pseudo-labeling result.

The processor 110 does not proceed with re-learning of the expert simulation model when the feedback-reflecting labeling result is less accurate than the pseudo-labeling result, but the processor 1100 determines that the feedback-reflecting labeling result is higher in accuracy than the pseudo-labeling result. Re-learning of the simulated expert model may be performed In this case, the processor 110 may re-train the simulated expert model using the feedback-reflected labeling result.

By configuring the training data as described above, it is possible to sufficiently provide the evidence data for labeling when the model is trained, and to allow the model to output the evidence data for labeling. Thereby, it is possible to provide sufficient data for training of the labeling model, and to obtain sufficient interpretation grounds for the labeling inference result of the model.

In addition, the processor 110 may generate feedback information with respect to data inferred by the learned model, apply it, and reconstruct the training data. Through this, the learning data can be further advanced. By continuously improving the learning data, the expert simulation model can perform a labeling task at a higher level, and can provide a higher level of labeling judgment basis information.

Referring to FIG. 7 , the processor may extract one or more identifiable objects from the image data ( S310 ).

The image data may include a plurality of identifiable objects. The processor 110 may extract a plurality of identifiable objects from the image data. In this case, the extraction of the object may be performed using object detection and localization techniques in the computer vision area.

The processor may infer one or more metadata and label information for each of one or more identifiable objects included in the image data (S320).

8 is a simplified, general schematic diagram of an exemplary computing environment in which embodiments of the present disclosure may be implemented.

Although the present disclosure has been described above generally as being capable of being implemented by a computing device, those skilled in the art will appreciate that the present disclosure is a combination of hardware and software and/or in combination with computer-executable instructions and/or other program modules that may be executed on one or more computers. It will be appreciated that it can be implemented as a combination.

Generally, program modules include routines, programs, components, data structures, etc. that perform particular tasks or implement particular abstract data types. In addition, those skilled in the art will appreciate that the methods of the present disclosure are suitable for single-processor or multiprocessor computer systems, minicomputers, mainframe computers as well as personal computers, handheld computing devices, microprocessor-based or programmable consumer electronics, etc. (each of which is It will be appreciated that other computer system configurations may be implemented, including those that may operate in connection with one or more associated devices.

The described embodiments of the present disclosure may also be practiced in distributed computing environments where certain tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote memory storage devices.

Computers typically include a variety of computer-readable media. Any medium accessible by a computer can be a computer readable medium, and such computer readable media includes volatile and nonvolatile media, transitory and non-transitory media, removable and non-transitory media. including removable media. By way of example, and not limitation, computer-readable media may include computer-readable storage media and computer-readable transmission media. Computer-readable storage media includes volatile and non-volatile media, transitory and non-transitory media, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. includes media. A computer-readable storage medium may be RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital video disk (DVD) or other optical disk storage device, magnetic cassette, magnetic tape, magnetic disk storage device, or other magnetic storage device. device, or any other medium that can be accessed by a computer and used to store the desired information.

Computer readable transmission media typically embodies computer readable instructions, data structures, program modules or other data, etc. in a modulated data signal such as a carrier wave or other transport mechanism, and Includes any information delivery medium. The term modulated data signal means a signal in which one or more of the characteristics of the signal is set or changed so as to encode information in the signal. By way of example, and not limitation, computer-readable transmission media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared, and other wireless media. Combinations of any of the above are also intended to be included within the scope of computer-readable transmission media.

An example environment 1100 implementing various aspects of the disclosure is shown including a computer 1102 , the computer 1102 including a processing unit 1104 , a system memory 1106 , and a system bus 1108 . do. The system bus 1108 couples system components, including but not limited to system memory 1106 , to the processing device 1104 . The processing device 1104 may be any of a variety of commercially available processors. Dual processor and other multiprocessor architectures may also be used as processing unit 1104 .

The system bus 1108 may be any of several types of bus structures that may further be interconnected to a memory bus, a peripheral bus, and a local bus using any of a variety of commercial bus architectures. System memory 1106 includes read only memory (ROM) 1110 and random access memory (RAM) 1112 . A basic input/output system (BIOS) is stored in non-volatile memory 1110, such as ROM, EPROM, EEPROM, etc., which is the basic input/output system (BIOS) that helps transfer information between components within computer 1102, such as during startup. contains routines. RAM 1112 may also include high-speed RAM, such as static RAM, for caching data.

The computer 1102 may also be configured for external use within an internal hard disk drive (HDD) 1114 (eg, EIDE, SATA) - this internal hard disk drive 1114 may also be configured for external use within a suitable chassis (not shown). Yes-, magnetic floppy disk drive (FDD) 1116 (eg, for reading from or writing to removable diskette 1118), and optical disk drive 1120 (eg, CD-ROM) for reading from, or writing to, disk 1122, or other high capacity optical media such as DVD. The hard disk drive 1114 , the magnetic disk drive 1116 , and the optical disk drive 1120 are connected to the system bus 1108 by the hard disk drive interface 1124 , the magnetic disk drive interface 1126 , and the optical drive interface 1128 , respectively. ) can be connected to The interface 1124 for external drive implementation includes at least one or both of Universal Serial Bus (USB) and IEEE 1394 interface technologies.

These drives and their associated computer-readable media provide non-volatile storage of data, data structures, computer-executable instructions, and the like. In the case of computer 1102, drives and media correspond to storing any data in a suitable digital format. Although the description of computer readable media above refers to HDDs, removable magnetic disks, and removable optical media such as CDs or DVDs, those skilled in the art will use zip drives, magnetic cassettes, flash memory cards, cartridges, etc. It will be appreciated that other tangible computer-readable media such as etc. may also be used in the exemplary operating environment and any such media may include computer-executable instructions for performing the methods of the present disclosure.

A number of program modules may be stored in the drive and RAM 1112 , including an operating system 1130 , one or more application programs 1132 , other program modules 1134 , and program data 1136 . All or portions of the operating system, applications, modules, and/or data may also be cached in RAM 1112 . It will be appreciated that the present disclosure may be implemented in various commercially available operating systems or combinations of operating systems.

A user may enter commands and information into the computer 1102 via one or more wired/wireless input devices, for example, a pointing device such as a keyboard 1138 and a mouse 1140 . Other input devices (not shown) may include a microphone, IR remote control, joystick, game pad, stylus pen, touch screen, and the like. Although these and other input devices are often connected to the processing unit 1104 through an input device interface 1142 that is connected to the system bus 1108, parallel ports, IEEE 1394 serial ports, game ports, USB ports, IR interfaces, It may be connected by other interfaces, etc.

A monitor 1144 or other type of display device is also coupled to the system bus 1108 via an interface, such as a video adapter 1146 . In addition to the monitor 1144, the computer typically includes other peripheral output devices (not shown), such as speakers, printers, and the like.

Computer 1102 may operate in a networked environment using logical connections to one or more remote computers, such as remote computer(s) 1148 via wired and/or wireless communications. Remote computer(s) 1148 may be workstations, computing device computers, routers, personal computers, portable computers, microprocessor-based entertainment devices, peer devices, or other common network nodes, and are typically connected to computer 1102 . Although it includes many or all of the components described for it, only memory storage device 1150 is shown for simplicity. The logical connections shown include wired/wireless connections to a local area network (LAN) 1152 and/or a larger network, eg, a wide area network (WAN) 1154 . Such LAN and WAN networking environments are common in offices and companies, and facilitate enterprise-wide computer networks, such as intranets, all of which can be connected to a worldwide computer network, for example, the Internet.

When used in a LAN networking environment, the computer 1102 is connected to the local network 1152 through a wired and/or wireless communication network interface or adapter 1156 . Adapter 1156 may facilitate wired or wireless communication to LAN 1152 , which also includes a wireless access point installed therein for communicating with wireless adapter 1156 . When used in a WAN networking environment, the computer 1102 may include a modem 1158, be connected to a communication computing device on the WAN 1154, or establish communications over the WAN 1154, such as over the Internet. have other means. A modem 1158 , which may be internal or external and a wired or wireless device, is coupled to the system bus 1108 via a serial port interface 1142 . In a networked environment, program modules described for computer 1102 , or portions thereof, may be stored in remote memory/storage device 1150 . It will be appreciated that the network connections shown are exemplary and other means of establishing a communication link between the computers may be used.

Computer 1102 may be associated with any wireless device or object that is deployed and operates in wireless communication, for example, printers, scanners, desktop and/or portable computers, portable data assistants (PDAs), communications satellites, wireless detectable tags. It operates to communicate with any device or place, and phone. This includes at least Wi-Fi and Bluetooth wireless technologies. Accordingly, the communication may be a predefined structure as in a conventional network or may simply be an ad hoc communication between at least two devices.

Wi-Fi (Wireless Fidelity) makes it possible to connect to the Internet, etc. without a wire. Wi-Fi is a wireless technology such as cell phones that allows these devices, eg, computers, to transmit and receive data indoors and outdoors, ie anywhere within range of a base station. Wi-Fi networks use a radio technology called IEEE 802.11 (a, b, g, etc.) to provide secure, reliable, and high-speed wireless connections. Wi-Fi can be used to connect computers to each other, to the Internet, and to wired networks (using IEEE 802.3 or Ethernet). Wi-Fi networks may operate in unlicensed 2.4 and 5 GHz radio bands, for example, at 11 Mbps (802.11a) or 54 Mbps (802.11b) data rates, or in products that include both bands (dual band). .

One of ordinary skill in the art of this disclosure will understand that information and signals may be represented using any of a variety of different technologies and techniques. For example, data, instructions, instructions, information, signals, bits, symbols and chips that may be referenced in the above description are voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields particles or particles, or any combination thereof.

Those of ordinary skill in the art of the present disclosure will recognize that the various illustrative logical blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein include electronic hardware, (convenience For this purpose, it will be understood that it may be implemented by various forms of program or design code (referred to herein as software) or a combination of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. A person skilled in the art of the present disclosure may implement the described functionality in various ways for each specific application, but such implementation decisions should not be interpreted as a departure from the scope of the present disclosure.

The various embodiments presented herein may be implemented as methods, apparatus, or articles of manufacture using standard programming and/or engineering techniques. The term article of manufacture includes a computer program, carrier, or media accessible from any computer-readable storage device. For example, computer-readable storage media include magnetic storage devices (eg, hard disks, floppy disks, magnetic strips, etc.), optical disks (eg, CDs, DVDs, etc.), smart cards, and flash drives. memory devices (eg, EEPROMs, cards, sticks, key drives, etc.). Also, various storage media presented herein include one or more devices and/or other machine-readable media for storing information.

It is understood that the specific order or hierarchy of steps in the presented processes is an example of exemplary approaches. Based on design priorities, it is understood that the specific order or hierarchy of steps in the processes may be rearranged within the scope of the present disclosure. The appended method claims present elements of the various steps in a sample order, but are not meant to be limited to the specific order or hierarchy presented.

The description of the presented embodiments is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Thus, the present disclosure is not intended to be limited to the embodiments presented herein, but is to be construed in the widest scope consistent with the principles and novel features presented herein.

As described above, related contents have been described in the best mode for carrying out the invention.

Claims

A computer program stored on a computer-readable storage medium, the computer program comprising instructions that, when executed on one or more processors of a computer device, cause the following steps to be performed for data processing, the steps comprising:

recognizing a training data set including one or more metadata and label information for each training data, the metadata being information on which the label information is a basis for giving the training data; and

training an expert simulation model based on the training data set;

containing,

A computer program stored in a computer-readable storage medium.
The method of claim 1,

The metadata is

When the learning data includes an identifiable object, including object region definition information of the identifiable object,

A computer program stored in a computer-readable storage medium.
3. The method of claim 2,

The object region definition information of the identifiable object includes:

including the shape information of the object area and parameter value information according to the shape information of the object area,

A computer program stored in a computer-readable storage medium.
The method of claim 1,

The metadata is

Including labeling behavior information recognized in the labeling process,

A computer program stored in a computer-readable storage medium.
5. The method of claim 4,

The labeling behavior information is

Label correction information, object region movement information, object region shape change information or object region rotation information, zoom-in behavior information, zoom-out behavior information, and contrast such as adjusting color to black and white Containing at least one of the contrast adjustment behavior information for

A computer program stored in a computer-readable storage medium.
The method of claim 1,

Each of the one or more metadata,

further containing time information;

A computer program stored in a computer-readable storage medium.
The method of claim 1,

The expert simulation model is,

learned to infer at least one of one or more metadata about an identifiable object or label information of the identifiable object,

A computer program stored in a computer-readable storage medium.
The method of claim 1,

performing pseudo labeling from image data using the trained expert simulation model;

further comprising,

A computer program stored in a computer-readable storage medium.
9. The method of claim 8,

The step of performing the pseudo-labeling comprises:

extracting one or more identifiable objects from the image data; and

inferring one or more metadata and label information for each of the one or more identifiable objects included in the image data;

containing,

A computer program stored in a computer-readable storage medium.
10. The method of claim 9,

Each of the one or more metadata for each of the identifiable objects inferred by the expert simulation model,

Includes labeling behavior information and time information inferred in the pseudo-labeling process,

Labeling behavior information and time information inferred in the pseudo labeling process are generated based on one or more metadata about the identifiable object collected up to a previous time,

A computer program stored in a computer-readable storage medium.
9. The method of claim 8,

receiving feedback information on the pseudo-labeling result; and

retraining the expert simulation model based on the feedback information;

further comprising,

A computer program stored in a computer-readable storage medium.
Memory; and

processor;

including,

The processor is

Recognizing a training data set including metadata and label information for each training data, wherein the metadata is information on which the label information is a basis for giving the training data;

training an expert simulation model based on the training data set,

Professional simulated computing device.
A computer-readable recording medium storing a data structure corresponding to a parameter of a neural network that is at least partially updated in a learning process, wherein the operation of the neural network is based at least in part on the parameter, the learning process comprising:

recognizing a training data set including metadata and label information for each of the training data, the metadata being information on which the label information is a basis for giving the training data; and

training an expert simulation model based on the training data set;

containing,

A computer-readable recording medium having a data structure stored thereon.
A method for training an expert simulation model performed on one or more processors of a computing device, the method comprising:

recognizing a training data set including metadata and label information for each of the training data, the metadata being information on which the label information is a basis for giving the training data; and

training an expert simulation model based on the training data set;

containing,

A method for training an expert simulation model performed on one or more processors of a computing device.