KR102670085B1 - Method for predicting product defects using neural network model - Google Patents

Abstract

According to one embodiment of the present disclosure, a method for predicting whether an article is defective is disclosed, which is performed by one or more processors of a computing device.
The method includes obtaining judgment target data; Inputting the judgment target data into a neural network model and extracting a first feature map; To extract a first detailed feature map to extract features related to whether there is a defect based on the first feature map, and to extract features of a different size from the first detailed feature map based on the first feature map. extracting a second detailed feature map; extracting a second feature map based on the first detailed feature map and the second detailed feature map; And it may include predicting whether there is a defect based on the extracted second feature map.

Description

A method for predicting whether a product is defective using a neural network model {METHOD FOR PREDICTING PRODUCT DEFECTS USING NEURAL NETWORK MODEL}

The present disclosure relates to a method of predicting whether a product is defective using a neural network model, and more specifically, using a neural network model to determine which category among a plurality of categories the judgment target data belongs to, and whether the judgment target data is defective. It is about a method for predicting.

Because an efficient quality monitoring system can prevent inconvenience and additional costs due to defects in products, automated surface defect inspection can be widely used in industrial manufacturing processes. However, various industrial products had different shapes, materials, and sizes, and a single industrial vision inspection system could not handle all tasks due to the complexity and diversity of industrial products. Accordingly, each factory product required its own inspection system, but when quality was previously inspected using human workers to inspect goods for defects, it was necessary to perform a high-level detailed inspection process during the inspection time due to eye fatigue. This was difficult, and required breaks at certain times to continue performing accurate work, and each worker had to have sufficient experience in analyzing items to ensure quality inspection. Therefore, there was a problem of decreased production efficiency when using human workers to inspect quality.

Therefore, considering the high price, long processing time, and quality inspection consistency issues that occurred in the existing product defect inspection, a method is needed to ensure consistency of quality inspection while reducing the processing time and cost of defect inspection. And new technologies that can solve these problems or shortcomings are required.

Meanwhile, the present disclosure has been derived based at least on the technical background examined above, but the technical problem or purpose of the present disclosure is not limited to solving the problems or shortcomings examined above. In other words, in addition to the technical issues discussed above, the present disclosure can cover various technical issues related to the content to be described below.

The present disclosure uses a neural network model to determine which category the judgment target data belongs to among a plurality of categories and to predict whether the judgment target data is defective.

Meanwhile, the technical problem to be achieved by the present disclosure is not limited to the technical problems mentioned above, and may include various technical problems within the scope of what is apparent to those skilled in the art from the contents described below.

A method performed by a computing device according to an embodiment of the present disclosure for realizing the above-described problem is disclosed. The method includes obtaining judgment target data; Inputting the judgment target data into a neural network model and extracting a first feature map; extracting a first detailed feature map for obtaining features related to whether or not a defect is present based on the first feature map, and obtaining features of a different size from the first detailed feature map based on the first feature map. extracting a second detailed feature map; Obtaining a second feature map based on the first detailed feature map and the second detailed feature map; And it may include predicting whether there is a defect based on the obtained second feature map.

Alternatively, the determination target data may include at least one of a plurality of defective data or normal data consisting of a plurality of categories.

Alternatively, the judgment object data includes one of a plurality of judgment object data, and the step of inputting the judgment object data into the neural network model and extracting the first feature map includes the judgment object data having different sizes. adjusting the size of the judgment target data based on the sizes of a plurality of judgment target data; And it may include inputting the adjusted judgment target data into the neural network model.

Alternatively, adjusting the size of the judgment target data based on the sizes of the plurality of judgment target data comprised of different sizes may include calculating an average aspect ratio of the plurality of judgment target data; and adjusting the size of the judgment target data based on the calculated average aspect ratio.

Alternatively, the method may further include patching the size-adjusted determination target data.

Alternatively, the neural network model may include one or more layers and one or more blocks.

Alternatively, the first layer and the second layer may include convolution layers, and the number of filters in the first layer may be smaller than the number of filters in the second layer.

Alternatively, a first block of the one or more blocks included in the neural network model includes a first sub-layer and a second sub-layer, and the first sub-layer and the second sub-layer have different numbers of filters, and , the first sub-layer and the second sub-layer can be configured in parallel.

Alternatively, the step of inputting the judgment target data into the neural network model and extracting the first feature map includes inputting the judgment target data into a first layer among one or more layers included in the neural network model and , Extracting a first feature map, extracting a first detailed feature map for obtaining features related to defects based on the first feature map, and extracting the first detailed feature map based on the first feature map. The step of extracting a second detailed feature map to obtain features of a different size from the detailed feature map includes inputting the first feature map to a first block among one or more blocks included in the neural network model, It may include extracting the first detailed feature map and the second detailed feature map.

Alternatively, obtaining a second feature map based on the first detailed feature map and the second detailed feature map may include concatenating the first detailed feature map and the second detailed feature map. It may include obtaining a second feature map.

Alternatively, predicting whether there is a defect based on the obtained second feature map may include determining a category in which the judgment target data will be included based on the obtained second feature map; And it may include predicting whether the data to be determined is defective based on the determined category.

According to an embodiment of the present disclosure for realizing the above-described object, a computer program stored in a computer-readable storage medium is disclosed. When the computer program is executed on one or more processors, it causes the one or more processors to perform operations for predicting whether a product is defective, and the operations include: acquiring judgment target data; Inputting the judgment target data into a neural network model and extracting a first feature map; extracting a first detailed feature map for obtaining features related to whether or not a defect is present based on the first feature map, and obtaining features of a different size from the first detailed feature map based on the first feature map. extracting a second detailed feature map; Obtaining a second feature map based on the first detailed feature map and the second detailed feature map; and predicting whether there is a defect based on the obtained second feature map.

Alternatively, the judgment target data includes one of a plurality of judgment target data, and the operation of inputting the judgment target data into the neural network model and extracting the first feature map includes the judgment target data having different sizes. Adjusting the size of the judgment target data based on the sizes of a plurality of judgment target data; and inputting the adjusted judgment target data into the neural network model.

Alternatively, the operation of adjusting the size of the judgment target data based on the sizes of the plurality of judgment target data comprised of the different sizes may include calculating an average aspect ratio of the plurality of judgment target data; and adjusting the size of the determination target data based on the calculated average aspect ratio.

Alternatively, the operation may further include patching the size-adjusted determination target data.

Alternatively, the operation of inputting the judgment target data into the neural network model and extracting the first feature map includes inputting the judgment target data into a first layer among one or more layers included in the neural network model and , comprising the operation of extracting a first feature map, extracting a first detailed feature map for obtaining features related to defects based on the first feature map, and extracting the first detailed feature map based on the first feature map. 1 The operation of extracting a second detailed feature map to obtain features of a different size from the detailed feature map includes inputting the first feature map to a first block among one or more blocks included in the neural network model, It may include extracting the first detailed feature map and the second detailed feature map.

Alternatively, the operation of obtaining a second feature map based on the first detailed feature map and the second detailed feature map may include concatenating the first detailed feature map and the second detailed feature map. An operation of acquiring a second feature map may be included.

Alternatively, the operation of predicting whether there is a defect based on the obtained second feature map may include: determining a category in which the judgment target data will be included based on the obtained second feature map; and predicting whether the data to be determined is defective based on the determined category.

A computing device according to an embodiment of the present disclosure for realizing the above-described problem is disclosed. The device includes at least one processor; and a memory, wherein the processor acquires judgment target data; Input the judgment target data into a neural network model and extract a first feature map; extracting a first detailed feature map for obtaining features related to whether or not a defect is present based on the first feature map, and obtaining features of a different size from the first detailed feature map based on the first feature map. extract a second detailed feature map; Obtain a second feature map based on the first detailed feature map and the second detailed feature map; And it may be configured to predict whether there is a defect based on the obtained second feature map.

A data structure included in a computer-readable storage medium according to an embodiment of the present disclosure for realizing the above-described problem is disclosed. The data structure corresponds to parameters of a neural network, and the neural network performs the following steps based at least in part on the parameters, the steps comprising: obtaining judgment target data; Inputting the judgment target data into a neural network model and extracting a first feature map; To extract a first detailed feature map for obtaining features related to whether there is a defect based on the first feature map, and to obtain features of a different size from the first detailed feature map based on the first feature map. extracting a second detailed feature map; Obtaining a second feature map based on the first detailed feature map and the second detailed feature map; And it may include predicting whether there is a defect based on the obtained second feature map.

The present disclosure uses a neural network model to determine which category the judgment target data belongs to among a plurality of categories and to predict whether the judgment target data has a defect.

Meanwhile, the effects of the present disclosure are not limited to the effects mentioned above, and various effects may be included within the range apparent to those skilled in the art from the contents described below.

1 is a block diagram of a computing device for predicting whether a product is defective using a neural network model according to an embodiment of the present disclosure.
Figure 2 is a schematic diagram showing a network function according to an embodiment of the present disclosure.
Figure 3 is a flowchart showing a method for predicting whether a product is defective using a neural network model according to an embodiment of the present disclosure.
FIG. 4A is a schematic diagram illustrating a plurality of judgment target data composed of different sizes according to an embodiment of the present disclosure.
FIG. 4B is a schematic diagram illustrating a process for adjusting the size of judgment target data before inputting it into a neural network model according to an embodiment of the present disclosure.
Figure 5 is a schematic diagram illustrating the structure of a neural network model for predicting whether a product is defective according to an embodiment of the present disclosure.
FIG. 6 is a schematic diagram illustrating the structure of one or more blocks included in a neural network model according to an embodiment of the present disclosure.
7 is a brief, general schematic diagram of an example 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 disclosure. However, it is clear that these embodiments may be practiced without these specific descriptions.

As used herein, the terms “component,” “module,” “system,” and the like refer to a computer-related entity, hardware, firmware, software, a combination of software and hardware, or an implementation of software. For example, a component may be, but is not limited to, 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 can 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. Additionally, these components can execute from various computer-readable media having various data structures stored thereon. Components may transmit signals, for example, with one or more data packets (e.g., data and/or signals from one component interacting with other components in a local system, a distributed system, to other systems and over a network such as the Internet). Depending on the data being transmitted, they may communicate through local and/or remote processes.

Additionally, the term “or” is intended to mean an inclusive “or” and not an exclusive “or.” That is, unless otherwise specified or clear from context, “X utilizes A or B” is intended to mean one of the natural implicit substitutions. That is, either X uses A; X uses B; Or, if X uses both A and B, “X uses A or B” can apply to either of these cases. Additionally, the term “and/or” as used herein should be understood to refer to and include all possible combinations of one or more of the related listed items.

Additionally, the terms “comprise” and/or “comprising” should be understood to mean that the corresponding feature and/or element is present. However, the terms “comprise” and/or “comprising” should be understood as not excluding the presence or addition of one or more other features, elements and/or groups thereof. Additionally, unless otherwise specified or the context is clear to indicate a singular form, the singular terms herein and in the claims should generally be construed to mean “one or more.”

And, the term “at least one of A or B” should be interpreted to mean “a case containing only A,” “a case containing only B,” and “a case of combining A and B.”

Those skilled in the art will additionally understand that the various illustrative logical blocks, configurations, modules, circuits, means, logic, and algorithm steps described in connection with the embodiments disclosed herein may be implemented in electronic hardware, computer software, or both. It should be recognized that it can be implemented with combinations of. To clearly illustrate the 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 in hardware or software will depend on the specific application and design constraints imposed on the overall system. A skilled technician can implement the described functionality in a variety of ways for each specific application. However, such implementation decisions should not be construed as causing a departure from the scope of the present disclosure.

The description of the presented embodiments is provided to enable anyone 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. The general principles defined herein may be applied to other embodiments without departing from the scope of the disclosure. Therefore, the present invention is not limited to the embodiments presented herein. The present invention is to be interpreted in the broadest scope consistent with the principles and novel features presented herein.

In this disclosure, network function, artificial neural network, and neural network may be used interchangeably.

1 is a block diagram of a computing device for predicting whether a product is defective using a neural network model according to an embodiment of the present disclosure.

The configuration of the computing device 100 shown in FIG. 1 is only a simplified example. In one embodiment of the present disclosure, the computing device 100 may include different 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, a memory 130, and a network unit 150.

The processor 110 may be composed of one or more cores, and may include a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), and a tensor processing unit (TPU) of a computing device. unit) 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 a neural network model. The processor 110 processes input data for learning in deep learning (DL), extracts features from input data, calculates errors, and learns neural network models such as updating the weights of the neural network model using backpropagation. Calculations can be performed for . At least one of the CPU, GPGPU, and TPU of the processor 110 may process learning of the neural network model. For example, the CPU and GPGPU can work together to process neural network model learning and data classification using the neural network model. Additionally, in one embodiment of the present disclosure, the processors of a plurality of computing devices can be used together to process learning of a neural network model and data classification using the neural network model. Additionally, a computer program executed in a 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 130 may store any type of information generated or determined by the processor 110 and any type of information received by the network unit 150.

According to an embodiment of the present disclosure, the memory 130 is a flash memory type, hard disk type, multimedia card micro type, or card type memory (e.g. (e.g. SD or -Only Memory), and may include at least one type of storage medium among magnetic memory, magnetic disk, and optical disk. The computing device 100 may operate in connection with web storage that performs a storage function of the memory 130 on the Internet. The description of the memory described above is merely an example, and the present disclosure is not limited thereto.

The network unit 150 according to an embodiment of the present disclosure includes Public Switched Telephone Network (PSTN), x Digital Subscriber Line (xDSL), Rate Adaptive DSL (RADSL), Multi Rate DSL (MDSL), and VDSL ( A variety of wired communication systems can be used, such as Very High Speed DSL), Universal Asymmetric DSL (UADSL), High Bit Rate DSL (HDSL), and Local Area Network (LAN).

In addition, the network unit 150 presented in the present disclosure includes Code Division Multi Access (CDMA), Time Division Multi Access (TDMA), Frequency Division Multi Access (FDMA), Orthogonal Frequency Division Multi Access (OFDMA), and SC-FDMA ( A variety of wireless communication systems can be used, such as Single Carrier-FDMA) and other systems.

In the present disclosure, the network unit 150 may be configured regardless of the communication mode, such as wired or wireless, and may be composed of various communication networks such as a personal area network (PAN) and a wide area network (WAN). You can. Additionally, the network may be the well-known World Wide Web (WWW), and may also use wireless transmission technology used for short-distance communication, such as Infrared Data Association (IrDA) or Bluetooth. The techniques described in this disclosure can also be used in other networks mentioned above.

Figure 2 is a schematic diagram showing a network function 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 can generally consist of a set of interconnected computational units, which can be referred to as nodes. These nodes may also be referred to as neurons. A neural network consists of at least one node. Nodes (or neurons) that make up neural networks may be interconnected by one or more links.

Within a neural network, one or more nodes connected through a link may form a relative input node and output node relationship. The concepts of input node and output node are relative, and any node in an output node relationship with one node may be in an input node relationship with another node, and vice versa. As described above, input node to output node relationships can be created around links. One or more output nodes can be connected to one input node through a link, and vice versa.

In a relationship between an input node and an output node connected through one link, the value of the data of the output node may be determined based on the data input to the input node. Here, the link connecting the input node and the output node may have a weight. Weights may be variable and may be varied by the user or algorithm in order for the neural network to perform the desired function. For example, when one or more input nodes are connected to one output node by respective links, the output node is set to the values input to the input nodes connected to the output node and the links corresponding to each input node. The output node value can 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 output node relationship within the neural network. The characteristics of the neural network may be determined according to the number of nodes and links within the neural network, the correlation between the nodes and links, and the value of the weight assigned to each link. For example, if the same number of nodes and links exist and two neural networks with different weight values of the links exist, 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 that make up a neural network can form a layer. Some of the nodes constituting the neural network may form one layer based on the distances from the first input node. For example, a set of nodes with a distance n from the initial input node may constitute n layers. The distance from the initial input node can be defined by the minimum number of links that must be passed to reach the node from the initial input node. However, this definition of a layer is arbitrary for explanation purposes, and the order of a layer within a neural network may be defined in a different way than described above. For example, a layer of nodes may be defined by distance from the final output node.

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

The neural network according to an embodiment of the present disclosure is 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 it progresses from the input layer to the hidden layer. You 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 it progresses from the input layer to the hidden layer. there is. 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 it progresses from the input layer to the hidden layer. You can. A neural network according to another embodiment of the present disclosure may be a neural network that is a combination of the above-described neural networks.

A deep neural network (DNN) may refer to a neural network that includes multiple hidden layers in addition to the input layer and output layer. Deep neural networks allow you to identify latent structures in data. In other words, it is possible to identify the potential structure of a photo, text, video, voice, or music (e.g., what object is in the photo, what the content and emotion of the text are, what the content and emotion of the voice are, etc.) . Deep neural networks include convolutional neural networks (CNN), recurrent neural networks (RNN), auto encoders, generative adversarial networks (GAN), and restricted Boltzmann machines (RBM). machine), deep belief network (DBN), Q network, U network, Siamese network, Generative Adversarial Network (GAN), etc. The description of the deep neural network described above is only an example and the present disclosure is not limited thereto.

In one embodiment of the present disclosure, the network function may include an autoencoder. An autoencoder may be a type of artificial neural network to output output data similar to input data. The autoencoder may include at least one hidden layer, and an odd number of hidden layers may be placed between input and output layers. The number of nodes in each layer may be reduced from the number of nodes in the input layer to an intermediate layer called the bottleneck layer (encoding), and then expanded symmetrically and reduced from the bottleneck layer to the output layer (symmetrical to the input layer). Autoencoders can perform nonlinear dimensionality reduction. The number of input layers and output layers can be corresponded to the dimension after preprocessing of the input data. In an auto-encoder structure, the number of nodes in 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 located between the encoder and decoder) is too small, not enough information may be conveyed, so if it is higher than a certain number (e.g., more than half of the input layers, etc.) ) may be maintained.

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

Neural networks can be trained to minimize output errors. In neural network learning, learning data is repeatedly input into the neural network, the output of the neural network and the error of the target for the learning data are calculated, and the error of the neural network is transferred from the output layer of the neural network to the input layer in the direction of reducing the error. This is the process of updating the weight of each node in the neural network through backpropagation. In the case of teacher learning, learning data in which the correct answer is labeled in each learning data is used (i.e., labeled learning data), and in the case of non-teacher learning, the correct answer may not be labeled in each learning data. That is, for example, in the case of teacher learning regarding data classification, the learning data may be data in which each learning data is labeled with a category. Labeled training data is input to the neural network, and the 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 non-teachable learning for data classification, the error can be calculated by comparing the input training data with the neural network output. The calculated error is backpropagated in the reverse direction (i.e., from the output layer to the input layer) in the neural network, and the connection weight of each node in each layer of the neural network can be updated according to backpropagation. The amount of change in the connection weight of each updated node may be determined according to the learning rate. The neural network's calculation of input data and backpropagation of errors can constitute a learning cycle (epoch). The learning rate may be applied differently depending on the number of repetitions of the learning cycle of the neural network. For example, in the early stages of neural network training, a high learning rate can be used to increase efficiency by allowing the neural network to quickly achieve a certain level of performance, and in the later stages of training, a low learning rate can be used to increase accuracy.

In the learning of neural networks, the training data can generally be a subset of real data (i.e., the data to be processed using the learned neural network), and thus the error for the training data is reduced, but the error for the real data is reduced. There may be an incremental learning cycle. Overfitting is a phenomenon in which errors in actual data increase due to excessive learning on training data. For example, a phenomenon in which a neural network that learned a cat by showing a yellow cat fails to recognize that it is a cat when it sees a non-yellow cat may be a type of overfitting. Overfitting can cause errors in machine learning algorithms to increase. To prevent such overfitting, various optimization methods can be used. To prevent overfitting, methods such as increasing the learning data, regularization, dropout to disable some of the network nodes during the learning process, and use of a batch normalization layer can be applied. You can.

Figure 3 is a flowchart showing a method for predicting whether a product is defective using a neural network model according to an embodiment of the present disclosure.

The computing device 100 according to an embodiment of the present disclosure may directly obtain “information for predicting whether a product is defective” or receive it from an external system. The external system may be a server, database, etc. that stores and manages information for predicting whether a product is defective using a neural network model. The computing device 100 may use information obtained directly or received from an external system as “input data for predicting whether a product is defective.” Additionally, the neural network model may include a neural network model for predicting whether a product is defective. For example, the neural network model may include a classification model for surface defects of an article. However, the surface defect classification model is only an example, and the present disclosure is not limited to this and various examples may be included.

The computing device 100 may obtain judgment target data (S110). At this time, the judgment target data may include a plurality of defective data or normal data consisting of a plurality of categories. For example, the plurality of defect data may include surface defect image data and may be comprised of a plurality of categories such as scratches, cracks, dents, dents, and stains. However, surface defect image data and categories such as scratches, cracks, and nicks are only examples, and the data subject to judgment in the present disclosure is not limited to these and may include various examples. Meanwhile, the judgment target data can be input into a neural network model and used in the process of predicting whether there is a defect, and the judgment target data can be resized before being input into the neural network model. At this time, the judgment target data may include one of a plurality of judgment target data. Specifically, the computing device 100 may adjust the size of the judgment target data based on the sizes of the plurality of judgment target data composed of different sizes, and input the adjusted judgment target data into the neural network model. . For example, the computing device 100 may calculate an average aspect ratio of the plurality of judgment target data and adjust the size of the judgment target data based on the calculated average aspect ratio. Additionally, the computing device 100 may patch the size-adjusted determination target data, and a detailed description of this will be provided later with reference to FIGS. 4A and 4B.

The computing device 100 may input the judgment target data obtained through step S110 into the neural network model and extract the first feature map (S120). At this time, the neural network model may include one or more layers. Specifically, the one or more layers included in the neural network model include a first layer and a second layer, and the first layer and the second layer may be hierarchically configured according to the number of filters. For example, the first layer and the second layer include convolution layers, and the number of filters in the first layer may be smaller than the number of filters in the second layer. Specifically, the one or more layers included in the neural network model are structured hierarchically so that the number of filters in each layer gradually increases, so that the lower layer of the neural network model (i.e., a convolutional layer with a relatively small number of filters) includes edges, corners, etc. The same simple features are learned, and complex features such as texture, detailed shape, etc. can be learned in higher layers (i.e., convolutional layers with a relatively large number of filters). However, the one or more layers included in the neural network model are only examples and are not limited to this, and various types of layers may be included in the neural network model.

The computing device 100 extracts a first detailed feature map for obtaining features related to defects based on the first feature map obtained through step S120, and extracts the first detailed feature map based on the first feature map. A second detailed feature map may be extracted to obtain features of a different size from the feature map (S130). Meanwhile, the neural network model may include one or more blocks. At this time, a first block among the one or more blocks included in the neural network model includes a first sub-layer and a second sub-layer, the first sub-layer and the second sub-layer have different numbers of filters, and the first sub-layer The first sub-layer and the second sub-layer may be configured in parallel. For example, the one or more blocks may be composed of a plurality of convolution layers, each with a different kernel size, in parallel. Additionally, the computing device 100 inputs the first feature map into a first block among one or more blocks included in the neural network model and extracts the first detailed feature map and the second detailed feature map. You can. For example, the computing device 100 may input the first feature map into the first sub-layer included in the first block to extract the first detailed feature map, and the second detailed feature map included in the first block. The first feature map can be input to the sub-layer, and the second detailed feature map can be extracted based on it. Through this, the neural network model can obtain richer features needed in the process of predicting defects by parallelly acquiring features of various sizes and complexities included in the judgment target data. A detailed description is provided in FIG. 6 below. It is described later.

According to an embodiment of the present disclosure, the computing device 100 may obtain a second feature map based on the first detailed feature map and the second detailed feature map extracted through step S130 (S140). Specifically, the computing device 100 may concatenate the first detailed feature map and the second detailed feature map to obtain the second feature map. For example, the computing device 100 may input the first feature map into the first sub-layer included in the first block to extract the first detailed feature map, and the second detailed feature map included in the first block. The first feature map can be input to a sub layer, a second detailed feature map can be extracted based on this, and a second feature map can be obtained by concatenating the first feature map and the second feature map. You can. Additionally, the computing device 100 may input the second feature map to a pooling layer, and the pooling layer may include a maximum pooling layer. For example, the maximum pooling layer may have a pool size of 2x2 and a stride of 2. By inputting the second feature map to the maximum pooling layer, the computing device 100 can reduce the dimension of the feature map while maintaining the most important features of the data to be determined. A detailed description of this will be described later with reference to FIGS. 5 and 6.

According to an embodiment of the present disclosure, the computing device 100 may predict whether there is a defect based on the second feature map obtained through step S140 (S150). Specifically, the computing device 100 may determine a category in which the judgment target data will be included based on the second feature map, and may predict whether the judgment target data is defective based on the determined category. For example, the computing device 100 determines that the judgment target data is classified into one of a plurality of categories such as scratches, defect free, pits, imprint, roll mark, etc. You can decide to include it. Additionally, the computing device 100 may predict that the judgment target data is defective when the judgment target data is determined to be a category other than the defect free category. A detailed description of this will be provided below with reference to FIG. 5.

FIG. 4A is a schematic diagram illustrating a plurality of judgment target data composed of different sizes according to an embodiment of the present disclosure.

Referring to FIG. 4A, the computing device 100 may obtain a plurality of judgment target data 10-1 to 10-6 from which parts unnecessary for determining defects have been removed. Meanwhile, the plurality of judgment target data 10-1 to 10-6 may have different sizes. In addition, the plurality of judgment target data 10-1 to 10-6 may include a surface image of the product and may be composed of a plurality of categories such as scratches, cracks, dents, dents, and stains. However, the surface image of the product and categories such as scratches, cracks, and nicks are only examples, and the data subject to judgment in the present disclosure is not limited to this and may include various examples. For example, referring to FIG. 4A, the first judgment target data 10-1 has a width of 385, the second judgment target data 10-2 has a width of 570, and the third judgment target data 10-3 has a width of 570. Width: 686, fourth defect data (10-4) has a width of: 933, fifth judgment target data (10-5) has a width of: 1061, and sixth judgment target data (10-6) has a width of: 1575. The first to sixth judgment target data 10-1 to 10-6 may have the same vertical length. However, in order for the computing device 100 to accurately predict whether a product is defective using a neural network model, it is necessary to adjust the size of each judgment target data. At this time, the computing device 100 can reduce the memory requirements and computational costs required for learning the neural network model by reducing the size of the judgment target data 10-1 to 10-6. However, when the computing device 100 adjusts the judgment target data 10-1 to 10-6 to a size that is too small, the quality of the image deteriorates, making it difficult to detect some defects, and in the process of adjusting the judgment target data to a too small size, As defects are deleted, the defect prediction accuracy of the neural network model may decrease. Conversely, the computing device 100 uses the neural network model in a real-time defect prediction system because process costs may be high if the judgment target data 10-1 to 10-6 are adjusted to a size that is too large. It may not be suitable for this purpose. Therefore, in order to increase the defect prediction accuracy of the neural network model, data loss must be prevented during the process of adjusting the sizes of the plurality of judgment target data 10-1 to 10-6 of different sizes. To this end, a process for pre-processing the plurality of judgment target data 10-1 to 10-6 of different sizes may be performed, which will be described later with reference to FIG. 4B.

FIG. 4B is a schematic diagram illustrating a process for adjusting the size of judgment target data before inputting it into a neural network model according to an embodiment of the present disclosure.

Referring to FIG. 4B, the computing device 100 according to an embodiment of the present disclosure inputs one of the judgment target data 10-1 to 10-6 into a neural network model and utilizes it in the process of predicting whether there is a defect. The size of the judgment target data may be adjusted before being input to the neural network model. Specifically, the computing device 100 may calculate the average aspect ratio of the plurality of judgment target data 10-1 to 10-6 comprised of the plurality of categories. For example, referring to the first image 11 of FIG. 4B, the average aspect ratio of the plurality of judgment target data 10-1 to 10-6 composed of the plurality of categories can be calculated as 4.13:1. there is. Specifically, the average horizontal length of the plurality of judgment target data (10-1 to 10-6) can be calculated as (385+570+686+933+1061+1575)/6 = 868.3, and the average vertical length is 3584, so the average aspect ratio can be calculated as 3584: 868.3 = 4.13:1. Additionally, the computing device 100 may adjust the size of the data to be determined based on the calculated average aspect ratio. For example, referring to the resized judgment target data 12 of FIG. 4B, the computing device 100 maintains 4:1, which approximates the average aspect ratio of 4.13:1, while maintaining the plurality of defect data 10 The size can be adjusted from -1 to 10-6). Through this, the computing device 100 can preserve visual details of the plurality of defect data. Additionally, the computing device 100 may adjust the size by dividing the size-adjusted determination target data 12. Specifically, the computing device 100 may patch the size-adjusted determination target data 12. For example, referring to the patched decision target data 13 and the first patch 14 of the decision target data of FIG. 4B, the computing device 100 uses the resized decision target data 12 It can be divided into 16 image patches with an aspect ratio of 1:1, and the size of the first patch 14 of the judgment target data, which is one of the patched judgment target data 13, can be adjusted to 256x256. there is. Meanwhile, the computing device 100 patches the size-adjusted judgment target data 12 into 16 smaller-sized images and inputs them into the neural network model, predicting whether there is a defect, and thereby predicting whether the neural network model has a defect. can identify more accurate and detailed features, and the accuracy of defect prediction results can be improved. (Refer to the description of FIG. 6 below for experimental results.) However, in addition to the example of 4.13:1 calculated as the average aspect ratio, various aspect ratios may be calculated as the average aspect ratio. Meanwhile, the specific process by which the first patch 14 of the judgment target data is input to the neural network model to predict whether it is defective will be described later with reference to FIGS. 5 and 6.

Figure 5 is a schematic diagram illustrating the structure of a neural network model for predicting whether a product is defective according to an embodiment of the present disclosure.

Referring to FIG. 5, the neural network model 20 according to an embodiment of the present disclosure includes one or more layers (21-1 to 21-4) and one or more blocks (22-1 to 22-3). may include. At this time, the one or more layers (21-1 to 21-4) included in the neural network model include a first layer (21-1) and a second layer (21-2), and the first layer (21-2) 1) and the second layer 21-2 may be hierarchically configured according to the number of filters. At this time, the one or more layers (21-1 to 21-4) may be composed of convolution layers, and the number of filters of the first layer (21-1) is the same as that of the second layer (21-2). It may be smaller than the number of filters. Meanwhile, in Figures 5 and 6, FC means fully connected layer, f is the number of filters, k is the size of the kernel, s is stride, and p is padding. , kr may mean kernel regularization. In addition, terms such as first and second described throughout this specification are only meant to distinguish between components and are not limited thereto.

Referring to the example of FIG. 5, the process of the computing device 100 predicting whether or not there is a defect in the first patch 14 of the judgment target data of 256x256 size using the neural network model 20 is configured as follows. It can be.

1) The first patch 14 of the judgment target data with a size of 256x256 is placed in the first layer 21-1 where the number of filters is 32, the kernel size is 5x5, the stride is 1, and the padding is set to the same. You can input and apply batch normalization to the output to normalize the output and improve the stability and speed of learning. Additionally, after batch normalization is applied, the first feature map 30 can be extracted by passing it through a max pooling layer with a pool size of 2x2 and a stride of 2.

2) The computing device 100 inputs the first feature map 30 into the first block 22-1 included in the second structure of the neural network model, creating a first detailed feature map and a second detailed feature map. The second feature map 40 can be obtained by extracting, concatenating the first detailed feature map and the second detailed feature map and passing it through a max pooling layer with a pool size of 2x2 and a stride of 2. .

3) For the output of the second feature map 40, the number of filters is set to 128, the kernel size is 5x5, the stride is set to 1, and the padding is set to the same, and L2 kernel normalization is applied so that the model is not overloaded on the training data. It can be input to the second layer 21-2, which can prevent merging.

4) The output of the second layer 21-2 may be input to the second block 22-2 composed of detailed layers with the number of filters being 64, 64, 32, 96, 32, and 64, respectively.

5) The output of the second block 22-2 is the third layer (21-) where the number of filters is 256, the kernel size is 3x3, the stride is 1, the padding is set to the same, and L2 kernel normalization is applied. 3) can be entered.

6) The output of the third layer (21-3) may be input to the third block (22-3) composed of detailed layers with the number of filters being 128, 96, 96, 96, 64, and 128, respectively. The output of the third block can be input to the 4th 'layer 21-4 with the number of filters set to 512, the kernel size set to 3x3, the stride set to 1, the padding set to the same, and L2 kernel normalization applied. .

7) The output of the fourth layer 21-4 can be subjected to a flatten process that converts a multidimensional array into a one-dimensional array, and by passing through four FC (Fully connected layers), the category in which the judgment target data will be included can be determined. .

However, the processes 1) to 7) for predicting whether the first patch 14 of the judgment target data having a size of 256x256 is defective are only examples, and are not limited thereto, and are used in embodiments of the present disclosure. Accordingly, the presence or absence of defects in judgment target data of various sizes and shapes is predicted using a neural network model 20 including one or more layers (21-1 to 21-4) and one or more blocks (22-1 to 22-3). It can be. Meanwhile, a classification error may be calculated based on a result of predicting whether the judgment target data is defective, which is the output of the neural network model 20, a loss function may be calculated based on the calculated classification error, and the calculated loss. The neural network model may be trained such that the function is minimized. Specifically, the computing device 100 includes the judgment target data in one of a plurality of categories such as scratches, defect free, pits, imprint, and roll mark. You can decide that it is okay. Additionally, the computing device 100 may predict that the judgment target data is defective when the judgment target data is determined to be a category other than the defect free category. If the first patch 14 of the judgment target data is a scratch category and the neural network model 20 determines the first patch 14 of the judgment target data to be a pits category, classification An error can be calculated, a loss function can be calculated based on the calculated classification error, and the neural network model 20 can be trained such that the calculated loss function is minimized.

In this regard, according to another embodiment of the present disclosure, generation is performed using a generator network including a layer for extracting features of input data and a layer for enlarging the size of the input data. The neural network model can be trained by using the artificial defect data as the judgment target data. At this time, the artificial defect data may include data of multiple categories such as scratches, defect free, pits, imprint, and roll mark. At this time, the artificial defect data may be generated using a generation network including a layer for extracting features of the input data and a layer for enlarging the size of the input data. For example, the generation network may use a random noise vector to expand the size of the input data by using at least one layer (for example, a Conv2DTranspose layer) to expand the size of the input data. At least one layer (eg, Conv2D layer) for extracting features can be used to ensure that the features of the data are well expressed. Additionally, the generation network may include a layer for preventing artifacts (eg, a Conv2D layer with a filter number of 1), and the layer for preventing artifacts is one of a plurality of layers of the generation network. It can be located at the end, and by using a layer to prevent the artifacts in the output of the generation network, artifacts such as checkerboard patterns or other distortions can be prevented in the generated artificial defect data. Therefore, the computing device 100 generates artificial defect data using a generator network including a layer for extracting features of input data and a layer for enlarging the size of the input data. More realistic artificial defect data can be generated, and the generated artificial defect data can be usefully used as the judgment target data in the learning process of the neural network model for predicting defects. In addition, the generated artificial defect data is used as the judgment target data to reinforce the training set used in the learning process of the neural network model to predict defects, thereby mitigating problems such as overfitting, class bias, and data imbalance. You can.

Additionally, the neural network model 20 is structured hierarchically so that the number of filters in one or more layers (21-1 to 21-4) gradually increases, so that the lower layer of the neural network model 20 (i.e., the number of filters is relatively small) In the convolutional layer (21-1), it is learned to predict defects related to simple features such as edges, corners, etc., and in the upper layer (i.e., convolutional layer (21-4) with a relatively large number of filters), texture, detailed shapes, etc. It can be trained to predict defects related to complex features such as Accordingly, one or more layers (21-1 to 21-4) included in the neural network model 20 are hierarchically organized and learned according to the number of filters, thereby enabling more diverse features in each layer of the neural network model 20. Because it can learn, defects can be better predicted. However, the plurality of convolution layers included in the neural network model 20 are only examples, and the present invention is not limited thereto, and various types of layers may be included in the neural network model. In addition, the computing device 100 improves prediction accuracy by using the one or more blocks 22-1 to 22-3 included in the neural network model in the process of predicting whether the data to be determined is defective using the neural network model. can be increased, and a detailed description of this will be described later with reference to FIG. 6.

FIG. 6 is a schematic diagram illustrating the structure of one or more blocks included in a neural network model according to an embodiment of the present disclosure.

Referring to FIG. 6 for the structure of the one or more blocks 22-1 to 22-3, the first block 22-1 according to an embodiment of the present disclosure includes a plurality of detailed layers 23-1 to 22-3. 23-5, etc.), and the 1-1 detailed layer 23-1 and the second detailed layer 23-3 included in the first block 22-1 have different numbers of filters. , the 1-1 detailed layer 23-1 and the second detailed layer 23-3 may be configured in parallel. For example, the plurality of detailed layers included in the first block 22-1 are convolution layers or pooling layers (in the example of FIG. 6, the kernel size is 3x3, the stride is 1, and the padding is may mean the same maximum pooling layer). At this time, the max pooling layer divides the input feature map into non-overlapping rectangular regions and selects the maximum value within each region, thereby reducing the size of the resulting output feature map, reducing the computational complexity of the model and preventing overfitting. It can be. Meanwhile, f[0, 1, 2, … ,5] may mean the number of filters included in each detailed layer. However, f[0~5] is only an example and is not limited to this, and various examples can be used. Specifically, the number of filters included in the 1-1 detailed layer (23-1) is f[0], and the number of filters included in the 1-2 detailed layer (23-2) is f[1]. , the number of filters included in the second detailed layer 23-3 may be f[2]. For example, in the case of the first block 22-1 composed of detailed layers with the number of filters being 32, 16, 16, 32, 16, and 16, respectively, the 1-1 detailed layer 23-1 is a kernel The size of may be 3x3, the number of filters may be 32, the size of the kernel of the 1-2 detailed layer 23-2 may be 1x1, the number of filters may be 16, and the second detailed layer 23-2 may have a size of 1x1 and the number of filters may be 16. The layer 23-3 may have a kernel size of 1x1 and the number of filters may be 16, and the 3-1 detailed layer 23-4 may have a kernel size of 1x1 and the number of filters may be 32. The size of the kernel of the 3-2 detailed layer 23-5 may be 3x3, and the number of filters may be 16. Meanwhile, the computing device 100 inputs the first feature map 30 into the 1-1 detailed layer 23-1 and the 1-2 detailed layer 23-2 to create a first detailed feature map. Extract, input the first feature map 30 into the second detailed layer 23-3 to extract a second detailed feature map, and input the first feature map 30 into the 3-1 detailed layer. The third detailed feature map can be extracted by inputting it into the layer 23-4 and the 3-2 detailed layer 23-5. First, the computing device 100 determines the local features of the first feature map 30 by inputting the first feature map 30 into the 1-1 detailed layer 23-1 whose kernel size is 3x3. The output of the 1-1 detailed layer (23-1) is input to the 1-2 detailed layer (23-2) whose kernel size is 1x1 to obtain a first detailed feature map, thereby obtaining the first detailed feature map. Regional features of the feature map 30 and features between channels can be extracted. In addition, the computing device 100 may input the first feature map 30 to a maximum pooling layer to reduce the spatial size of the first feature map 30 and emphasize key features, By inputting the output of the maximum pooling layer to the second detailed layer 23-3 with a kernel size of 1x1 to obtain a second detailed feature map, features between channels of the first feature map 30 can be extracted. And since the number of channels is the same as the previous input, the amount of computation can be reduced. Additionally, the computing device 100 reduces the channel of the first feature map by using the 3-1 detailed layer 23-4 whose kernel size is 1x1 and then reduces the channel of the 3-2 detailed layer 23-4. By expanding again through -5), the amount of calculation required can be reduced. For example, if the computing device 100 uses only the 3-2 detailed layer 23-5 with a kernel size of 3x3, a 3x3 kernel is created by considering information between channels as well as local information. Since it can be expressed as the value of , one kernel must perform both roles. In contrast, if the 3-1 detailed layer 23-4 with a kernel size of 1x1 is used first, the 1x1 size kernel plays the role of adjusting the channels, so that features between channels can be extracted, and the following Since the kernel size of the 3-2 detailed layer 23-5 is 3x3, features can be extracted by focusing only on local information of the judgment target data. Accordingly, the computing device 100 extracts a third detailed feature map using the 3-1 detailed layer 23-4 first and then the 3-2 detailed layer 23-5, The roles of the two detailed layers (23-4 and 23-5) with different kernel sizes can be subdivided. In other words, the relationship information between channels is connected to parameters used in the 3-1 detailed layer 23-4 whose kernel size is 1x1, and the local information of the judgment target data is connected to the parameters used in the 3-1 detailed layer 23-4 whose kernel size is 3x3. Parameters used in the 3-2 detailed layer 23-5 may be connected to each other. Additionally, the computing device 100 may acquire the second feature map 40 by channel-wise concatenating the obtained first, second, and third detailed feature maps. At this time, the computing device 100 uses a plurality of detailed layers in parallel with different kernel sizes and different numbers of filters included in the one or more blocks, thereby providing visual characteristics of the judgment target data at various sizes and resolutions. A feature map (e.g., a second feature map 40, etc.) can be acquired, and through this feature map (e.g., a second feature map 40, etc.), the neural network model can be used at various scales and resolutions. Defect patterns and detailed information of the judgment target data can be obtained. In addition, the computing device 100 uses a feature map (eg, the second feature map 40, etc.) with visual characteristics of the acquired judgment target data in the process of predicting whether the judgment target data is defective. By doing so, the accuracy of prediction can be increased. In summary, the computing device 100 predicts whether the judgment target data is defective using the neural network model including one or more layers and one or more blocks, thereby providing defect patterns and details of the judgment target data at various scales and resolutions. can be obtained, and the accuracy of defect prediction results can be improved.

Meanwhile, the tables below compare existing defect prediction results with defect prediction results according to an embodiment of the present disclosure.

<Table 1: Performance of models with existing structures learned with existing judgment target data>

<Table 2: Performance of model with existing structure learned with resized judgment target data>

<Table 3: Performance of a neural network model configured according to an embodiment of the present disclosure learned with scaled judgment target data>

Comparing the results of Table 1, Table 2, and Table 3, compared to the prediction accuracy of 74% for the existing model learned using the judgment target data in Table 1 without adjusting the size, When the existing model is learned with the adjusted judgment target data in Table 2 and whether the judgment target data is defective, the prediction accuracy can be improved to 94%. Additionally, a neural network model composed of an embodiment of the present disclosure learned with the adjusted judgment target data of Table 3 is learned, and whether the determined target data is defective using the learned neural network model composed of an embodiment of the present disclosure. If is predicted, you can see that the prediction accuracy has improved to 98%. Therefore, the judgment target data is adjusted to an appropriate size according to the methods disclosed in the embodiments of the present disclosure, and the judgment target data is adjusted to an appropriate size using the adjusted judgment target data and a neural network model composed of the embodiments of the present disclosure. By predicting whether there is a defect, a technical effect can be obtained that can improve the accuracy of the defect prediction result.

<Table 4: Comparison of the performance of the neural network model of the present disclosure and the performance of existing models>

additionally. Referring to the results in Table 4, the neural network model of the present disclosure has a defect prediction accuracy of 98.46% for all categories of data, showing excellent performance compared to the remaining models (ResNet50, EfficientNetV2S, CNN-only, SCA model). there is. In particular, compared to the remaining models (ResNet50, EfficientNetV2S, CNN-only, SCA model), the accuracy of predicting whether the judgment target data in the scratch category is defective is 94.44% for the neural network model of this disclosure, which is 55.56% for ResNet50. It shows excellent performance compared to 50% accuracy of EfficientNetV2S and 77.78% accuracy of CNN-only and SCA. Therefore, according to the methods disclosed in the embodiments of the present disclosure, the judgment target data is adjusted to an appropriate size, and the judgment target data is adjusted to an appropriate size using the adjusted judgment target data and a neural network model composed of the embodiments of the present disclosure. By predicting whether there is a defect, a technical effect can be obtained that can improve the accuracy of the defect prediction result.

According to an embodiment of the present disclosure, a computer-readable medium storing a data structure is disclosed. Data structure can refer to the organization, management, and storage of data to enable efficient access and modification of data. Data structure can refer to the organization of data to solve a specific problem (e.g., retrieving data, storing data, or modifying data in the shortest possible time). A data structure may be defined as a physical or logical relationship between data elements designed to support a specific data processing function. Logical relationships between data elements may include connection relationships between user-defined data elements. Physical relationships between data elements may include actual relationships between data elements that are physically stored in a computer-readable storage medium (e.g., a persistent storage device). A data structure may specifically include a set of data, relationships between data, and functions or instructions applicable to the data. Effectively designed data structures allow computing devices to perform computations while minimizing the use of the computing device's resources. Specifically, computing devices can increase the efficiency of operations, reading, insertion, deletion, comparison, exchange, and search through effectively designed data structures.

Data structures can be divided into linear data structures and non-linear data structures depending on the type of data structure. A linear data structure may be a structure in which only one piece of data is connected to another piece of data. Linear data structures may include List, Stack, Queue, and Deque. A list can refer to a set of data that has an internal order. The list may include a linked list. A linked list may be a data structure in which data is connected in such a way that each data is connected in a single line with a pointer. In a linked list, a pointer may contain connection information to the next or previous data. Depending on its form, a linked list can be expressed as a singly linked list, a doubly linked list, or a circularly linked list. A stack may be a data listing structure that allows limited access to data. A stack can be a linear data structure in which data can be processed (for example, inserted or deleted) at only one end of the data structure. Data stored in the stack may have a data structure (LIFO-Last in First Out) where the later it enters, the sooner it comes out. A queue is a data listing structure that allows limited access to data. Unlike the stack, it can be a data structure (FIFO-First in First Out) where data stored later is released later. A deck can be a data structure that can process data at both ends of the data structure.

A non-linear data structure may be a structure in which multiple pieces of data are connected behind one piece of data. Nonlinear data structures may include graph data structures. A graph data structure can be defined by vertices and edges, and an edge can include a line connecting two different vertices. Graph data structure may include a tree data structure. A tree data structure may be a data structure in which there is only one path connecting two different vertices among a plurality of vertices included in the tree. In other words, 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. Below, it is described in a unified manner as a neural network. Data structures may include neural networks. And the data structure including the neural network may be stored in a computer-readable medium. Data structures including neural networks also include data preprocessed for processing by a neural network, data input to the neural network, weights of the neural network, hyperparameters of the neural network, data acquired from the neural network, activation functions associated with each node or layer of the neural network, neural network It may include a loss function for learning. A data structure containing a neural network may include any of the components disclosed above. In other words, 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 acquired from the neural network, activation functions associated with each node or layer of the neural network, neural network It may be configured to include all or any combination of the loss function for learning. In addition to the configurations described above, a data structure containing a neural network may include any other information that determines the characteristics of the neural network. Additionally, the data structure may include all types of data used or generated in the computational process of a 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 can generally consist of a set of interconnected computational units, which can be referred to as nodes. These nodes may also be referred to as neurons. A neural network consists of at least one node.

The data structure may include data input to the neural network. A data structure containing data input to a neural network may be stored in a computer-readable medium. Data input to the neural network may include learning data input during the neural network learning process and/or input data input to the neural network on which training has been completed. Data input to the neural network may include data that has undergone pre-processing and/or data subject to pre-processing. Preprocessing may include a data processing process to input data into a neural network. Therefore, the data structure may include data subject to preprocessing and data generated by preprocessing. The above-described data structure is only an example and the present disclosure is not limited thereto.

The data structure may include the weights of the neural network. (In this specification, weights and parameters may be used with the same meaning.) And the data structure including the weights of the neural network may be stored in a computer-readable medium. A neural network may include multiple weights. Weights may be variable and may be varied by the user or algorithm in order for the neural network to perform the desired function. For example, when one or more input nodes are connected to one output node by respective links, the output node is set to the values input to the input nodes connected to the output node and the links corresponding to each input node. Based on the weight, the data value output from the output node can be determined. The above-described data structure is only an example and the present disclosure is not limited thereto.

As an example and not a limitation, the weights may include weights that are changed during the neural network learning process and/or weights for which neural network learning has been completed. Weights that change during the neural network learning process may include weights that change at the start of the learning cycle and/or weights that change during the learning cycle. Weights for which neural network training has been completed may include weights for which a learning cycle has been completed. Therefore, the data structure including the weights of the neural network may include weights that are changed during the neural network learning process and/or the data structure including the weights for which neural network learning has been completed. Therefore, the above-mentioned weights and/or combinations of each weight are included in the data structure including the weights of the neural network. The above-described data structure is only 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 (e.g., memory, hard disk) after going through a serialization process. Serialization can be the process of converting a data structure into a form that can be stored on the same or a different computing device and later reorganized and used. Computing devices can transmit and receive data over a network by serializing data structures. Data structures containing the weights of a serialized neural network can be reconstructed on the same computing device or on a different 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 computational efficiency while minimizing the use of computing device resources (e.g., in non-linear data structures, B-Tree, Trie, m-way search tree, AVL tree, Red-Black Tree) may be included. The foregoing is merely an example and the present disclosure is not limited thereto.

The data structure may include hyper-parameters of a neural network. And the data structure including the hyperparameters of the neural network can be stored in a computer-readable medium. A hyperparameter may be a variable that can be changed by the user. Hyperparameters include, for example, learning rate, cost function, number of learning cycle repetitions, weight initialization (e.g., setting the range of weight values subject to weight initialization), Hidden Unit. It may include a number (e.g., number of hidden layers, number of nodes in hidden layers). The above-described data structure is only an example and the present disclosure is not limited thereto.

7 is a brief, general schematic diagram of an example computing environment in which embodiments of the present disclosure may be implemented.

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

Typically, program modules include routines, programs, components, data structures, etc. that perform specific tasks or implement specific abstract data types. Additionally, those skilled in the art will understand that the methods of the present disclosure can be used on single-processor or multiprocessor computer systems, minicomputers, mainframe computers, as well as personal computers, handheld computing devices, microprocessor-based or programmable consumer electronics, etc. It will be appreciated that other computer system configurations may be implemented, including, but not limited to, each of which may operate in connection with one or more associated devices.

The described embodiments of the disclosure can 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. Computer-readable media can be any medium that can be accessed by a computer, and such computer-readable media includes volatile and non-volatile media, transitory and non-transitory media, removable and non-transitory media. Includes 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 refers to volatile and non-volatile media, transient and non-transitory media, removable and non-removable, implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Includes media. Computer readable storage media may include RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital video disk (DVD) or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage. This includes, but is not limited to, a device, or any other medium that can be accessed by a computer and used to store desired information.

A computer-readable transmission medium typically implements computer-readable instructions, data structures, program modules, or other data on a modulated data signal, such as a carrier wave or other transport mechanism. Includes all information delivery media. The term modulated data signal refers to a signal in which one or more of the characteristics of the signal have been set or changed to encode information within 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 is shown that implements various aspects of the present disclosure, including a computer 1102, which includes a processing unit 1104, a system memory 1106, and a system bus 1108. do. System bus 1108 couples system components, including but not limited to system memory 1106, to processing unit 1104. Processing unit 1104 may be any of a variety of commercially available processors. Dual processors and other multiprocessor architectures may also be used as processing unit 1104.

System bus 1108 may be any of several types of bus structures that may further be interconnected to a memory bus, peripheral bus, and 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. The basic input/output system (BIOS) is stored in non-volatile memory 1110, such as ROM, EPROM, and EEPROM, and is a basic input/output system that helps transfer information between components within the computer 1102, such as during startup. Contains routines. RAM 1112 may also include high-speed RAM, such as static RAM, for caching data.

Computer 1102 may also include an internal hard disk drive (HDD) 1114 (e.g., EIDE, SATA)—the internal hard disk drive 1114 may also be configured for external use within a suitable chassis (not shown). Yes - a magnetic floppy disk drive (FDD) 1116 (e.g., for reading from or writing to a removable diskette 1118), and an optical disk drive 1120 (e.g., a CD-ROM for reading the disk 1122 or reading from or writing to other high-capacity optical media such as DVDs). Hard disk drive 1114, magnetic disk drive 1116, and optical disk drive 1120 are connected to system bus 1108 by hard disk drive interface 1124, magnetic disk drive interface 1126, and optical drive interface 1128, respectively. ) can be connected to. The interface 1124 for implementing an external drive 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. For computer 1102, drive and media correspond to storing any data in a suitable digital format. Although the description of computer-readable media above refers to removable optical media such as HDDs, removable magnetic disks, and CDs or DVDs, those of ordinary skill in the art would also recognize zip drives, magnetic cassettes, flash memory cards, and cartridges. It will be appreciated that other types of computer-readable media may also be used in the exemplary operating environment, and that any such media may contain 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 on various commercially available operating systems or combinations of operating systems.

A user may enter commands and information into computer 1102 through one or more wired/wireless input devices, such as a keyboard 1138 and a pointing device such as mouse 1140. Other input devices (not shown) may include microphones, IR remote controls, joysticks, game pads, stylus pens, touch screens, etc. These and other input devices are connected to the processing unit 1104 through an input device interface 1142, which is often connected to the system bus 1108, but may also include a parallel port, an IEEE 1394 serial port, a game port, a USB port, an IR interface, It can be connected by other interfaces, etc.

A monitor 1144 or other type of display device is also connected to system bus 1108 through an interface, such as a video adapter 1146. In addition to monitor 1144, computers typically include other peripheral output devices (not shown) such as speakers, printers, etc.

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 a workstation, computing device computer, router, personal computer, portable computer, microprocessor-based entertainment device, peer device, or other conventional network node, and is generally connected to computer 1102. For simplicity, only memory storage device 1150 is shown, although it includes many or all of the components described. The logical connections depicted include wired/wireless connections to a local area network (LAN) 1152 and/or a larger network, such as a wide area network (WAN) 1154. These 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, such as the Internet.

When used in a LAN networking environment, computer 1102 is connected to local network 1152 through wired and/or wireless communication network interfaces or adapters 1156. Adapter 1156 may facilitate wired or wireless communication to LAN 1152, which also includes a wireless access point installed thereon for communicating with wireless adapter 1156. When used in a WAN networking environment, the computer 1102 may include a modem 1158 or be connected to a communicating computing device on the WAN 1154 or to establish communications over the WAN 1154, such as over the Internet. Have other means. Modem 1158, which may be internal or external and a wired or wireless device, is coupled to system bus 1108 via 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 that other means of establishing a communications link between computers may be used.

Computer 1102 may be associated with any wireless device or object deployed and operating in wireless communications, such as a printer, scanner, desktop and/or portable computer, portable data assistant (PDA), communications satellite, wirelessly detectable tag. Performs actions to communicate with any device or location and telephone. This includes at least Wi-Fi and Bluetooth wireless technologies. Accordingly, communication may be a predefined structure as in a conventional network or may simply be ad hoc communication between at least two devices.

Wi-Fi (Wireless Fidelity) allows connection to the Internet, etc. without wires. Wi-Fi is a wireless technology, like cell phones, that allows these devices, such as computers, to send and receive data indoors and outdoors, anywhere within the coverage area of a cell tower. Wi-Fi networks use wireless 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 can operate in the unlicensed 2.4 and 5 GHz wireless bands, for example, at data rates of 11 Mbps (802.11a) or 54 Mbps (802.11b), or in products that include both bands (dual band). .

Those skilled in the art will understand that information and signals may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols and chips that may be referenced in the above description include voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields. It can be expressed by particles or particles, or any combination thereof.

Those skilled in the art will understand that the various illustrative logical blocks, modules, processors, means, circuits and algorithm steps described in connection with the embodiments disclosed herein may be used in electronic hardware, (for convenience) 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 interoperability of hardware and software, various illustrative components, blocks, modules, circuits and steps have been described above generally with respect to their functionality. Whether this functionality is implemented as hardware or software depends on the specific application and design constraints imposed on the overall system. A person skilled in the art of this disclosure may implement the described functionality in various ways for each specific application, but such implementation decisions should not be construed as departing from the scope of this disclosure.

The various embodiments presented herein may be implemented as a method, apparatus, or article 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 (e.g., hard disks, floppy disks, magnetic strips, etc.), optical disks (e.g., CDs, DVDs, etc.), smart cards, and flash. Includes, but is not limited to, memory devices (e.g., EEPROM, cards, sticks, key drives, etc.). Additionally, various storage media presented herein include one or more devices and/or other machine-readable media for storing information.

It is to be understood that the specific order or hierarchy of steps in the processes presented is an example of illustrative approaches. It is to be understood that the specific order or hierarchy of steps in processes may be rearranged within the scope of the present disclosure, based on design priorities. The appended method claims present elements of the various steps in a sample order but are not meant to be limited to the particular 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 apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the scope of the disclosure. Thus, the present disclosure is not limited to the embodiments presented herein but is to be interpreted in the broadest scope consistent with the principles and novel features presented herein.

Claims (22)

A method for predicting whether an article is defective, performed by a computing device, comprising:
Obtaining judgment target data;
adjusting the size of the judgment target data based on the size of a plurality of judgment target data composed of different sizes;
Inputting the judgment target data adjusted to the same size into a neural network model and extracting a first feature map;
extracting a first detailed feature map for obtaining features related to whether or not a defect is present based on the first feature map, and obtaining features of a different size from the first detailed feature map based on the first feature map. extracting a second detailed feature map;
Obtaining a second feature map based on the first detailed feature map and the second detailed feature map; and
predicting whether there is a defect based on the obtained second feature map;
Including,
method.
According to claim 1,
The data subject to the above judgment is,
Multiple defect data consisting of multiple categories or
Containing at least one of the normal data,
method.
According to claim 1,
The data subject to the above judgment is,
Containing one of the plurality of judgment target data,
method.
According to claim 3,
The step of adjusting the size of the judgment target data based on the size of the plurality of judgment target data composed of different sizes includes,
calculating an average aspect ratio of the plurality of judgment target data; and
adjusting the size of the judgment target data based on the calculated average aspect ratio;
Including,
method.
According to claim 4,
Patching the size-adjusted judgment target data;
Containing more,
method.
According to claim 1,
The neural network model is,
Containing one or more layers and one or more blocks,
method.
According to claim 6,
The one or more layers included in the neural network model are:
comprising a first layer and a second layer,
The first layer and the second layer are hierarchically organized according to the number of filters,
method.
According to claim 7,
The first layer and the second layer include a convolution layer,
The number of filters in the first layer is,
less than the number of filters in the second layer,
method.
According to claim 6,
The first block among the one or more blocks included in the neural network model is,
comprising a first sub-layer and a second sub-layer,
The first sub-layer and the second sub-layer have different numbers of filters,
The first sub-layer and the second sub-layer are configured in parallel,
method.
According to claim 6,
The step of inputting the judgment target data adjusted to the same size into the neural network model and extracting the first feature map is,
Inputting the judgment target data into a first layer among one or more layers included in the neural network model and extracting a first feature map,
extracting a first detailed feature map for obtaining features related to whether or not a defect is present based on the first feature map, and obtaining features of a different size from the first detailed feature map based on the first feature map. The step of extracting the second detailed feature map is,
Inputting the first feature map into a first block among one or more blocks included in the neural network model, and extracting the first detailed feature map and the second detailed feature map.
method.
According to claim 1,
Obtaining a second feature map based on the first detailed feature map and the second detailed feature map includes:
Obtaining the second feature map by concatenating the first detailed feature map and the second detailed feature map;
Including,
method.
According to claim 1,
The step of predicting whether there is a defect based on the obtained second feature map,
determining a category in which the judgment target data will be included based on the obtained second feature map; and
predicting whether the judgment target data is defective based on the determined category;
Including,
method.
A computer program stored in a computer-readable storage medium, wherein the computer program, when executed by one or more processors, causes the one or more processors to perform operations for predicting whether an article is defective, the operations comprising:
An operation of acquiring judgment target data;
An operation of adjusting the size of the judgment target data based on the sizes of a plurality of judgment target data composed of different sizes;
Inputting the judgment target data adjusted to the same size into a neural network model and extracting a first feature map;
extracting a first detailed feature map for obtaining features related to defects based on the first feature map, and obtaining features of a different size from the first detailed feature map based on the first feature map. extracting a second detailed feature map;
Obtaining a second feature map based on the first detailed feature map and the second detailed feature map; and
Predicting whether there is a defect based on the obtained second feature map;
Including,
A computer program stored on a computer-readable storage medium.
According to claim 13,
The data subject to the above judgment is,
Containing one of the plurality of judgment target data,
A computer program stored on a computer-readable storage medium.
According to claim 14,
The operation of adjusting the size of the judgment target data based on the sizes of the plurality of judgment target data composed of different sizes,
calculating an average aspect ratio of the plurality of judgment target data; and
Adjusting the size of the judgment target data based on the calculated average aspect ratio;
Including,
A computer program stored on a computer-readable storage medium.
According to claim 15,
An operation of patching the size-adjusted judgment target data;
Containing more,
A computer program stored on a computer-readable storage medium.
According to claim 13,
The neural network model is,
Containing one or more layers and one or more blocks,
A computer program stored on a computer-readable storage medium.
According to claim 17,
The operation of inputting the judgment target data adjusted to the same size into the neural network model and extracting the first feature map is,
Inputting the judgment target data into a first layer among one or more layers included in the neural network model and extracting a first feature map,
extracting a first detailed feature map for obtaining features related to defects based on the first feature map, and obtaining features of a different size from the first detailed feature map based on the first feature map. The operation of extracting the second detailed feature map is:
Including the operation of inputting the first feature map into a first block among one or more blocks included in the neural network model and extracting the first detailed feature map and the second detailed feature map.
A computer program stored on a computer-readable storage medium.
According to claim 13,
The operation of obtaining a second feature map based on the first detailed feature map and the second detailed feature map includes:
Obtaining the second feature map by concatenating the first detailed feature map and the second detailed feature map;
Including,
A computer program stored on a computer-readable storage medium.
According to claim 13,
The operation of predicting whether there is a defect based on the obtained second feature map,
determining a category in which the judgment target data will be included based on the obtained second feature map; and
predicting whether the judgment target data is defective based on the determined category;
Including,
A computer program stored on a computer-readable storage medium.
As a computing device,
at least one processor; and
Memory
Including,
The at least one processor,
Obtain data to be judged;
adjusting the size of the judgment target data based on the size of a plurality of judgment target data composed of different sizes;
Input the judgment target data adjusted to the same size into a neural network model and extract a first feature map;
extracting a first detailed feature map for obtaining features related to whether or not a defect is present based on the first feature map, and obtaining features of a different size from the first detailed feature map based on the first feature map. extract a second detailed feature map;
Obtain a second feature map based on the first detailed feature map and the second detailed feature map; and
configured to predict whether there is a defect based on the obtained second feature map,
Computing device.
A computer-readable storage medium storing parameters of a neural network, wherein the neural network performs the following steps based at least in part on the parameters, the steps comprising:
Obtaining judgment target data;
adjusting the size of the judgment target data based on the size of a plurality of judgment target data composed of different sizes;
Inputting the judgment target data adjusted to the same size into a neural network model and extracting a first feature map;
To extract a first detailed feature map for obtaining features related to defects based on the first feature map, and to obtain features of a different size from the first detailed feature map based on the first feature map. extracting a second detailed feature map;
Obtaining a second feature map based on the first detailed feature map and the second detailed feature map; and
predicting whether there is a defect based on the obtained second feature map;
Including,
Computer readable storage medium.