• Journal of Broadcast Engineering
• /
• v.27 no.3
• /
• pp.378-390
• /
• 2022

With the recent development of deep composite multiplication neural network learning, deep learning techniques applied to single-image super-resolution have shown good results, and the strong expression ability of deep networks has enabled complex nonlinear mapping between low-resolution and high-resolution images. However, there are limitations in applying it to real-time or low-power devices with increasing parameters and computational amounts due to excessive use of composite multiplication neural networks. This paper uses blocks that extract hierarchical characteristics little by little using information distillation and suggests the Recursive Distillation Super Resolution Network (RDSRN), a lightweight network that improves performance by making more accurate high frequency components through high frequency residual purification blocks. It was confirmed that the proposed network restores images of similar quality compared to RDN, restores images 3.5 times faster with about 32 times fewer parameters and about 10 times less computation, and produces 0.16 dB better performance with about 2.2 times less parameters and 1.8 times faster processing time than the existing lightweight network CARN.

• Convergence Security Journal
• /
• v.23 no.2
• /
• pp.47-54
• /
• 2023

The global popularity of K-content(Korean Wave) has led to a continuous increase in copyright infringement cases involving domestic works, not only within the country but also overseas. In response to this trend, there is active research on technologies for detecting illegal distribution sites of domestic copyrighted materials, with recent studies utilizing the characteristics of domestic illegal distribution sites that often include a significant number of advertising banners. However, the application of detection techniques similar to those used domestically is limited for overseas illegal distribution sites. These sites may not include advertising banners or may have significantly fewer ads compared to domestic sites, making the application of detection technologies used domestically challenging. In this study, we propose a detection technique based on the similarity comparison of links and text trees, leveraging the characteristic of including illegal sharing posts and images of copyrighted materials in a similar hierarchical structure. Additionally, to accurately compare the similarity of large-scale trees composed of a massive number of links, we utilize Graph Neural Network (GNN). The experiments conducted in this study demonstrated a high accuracy rate of over 95% in classifying regular sites and sites involved in the illegal distribution of copyrighted materials. Applying this algorithm to automate the detection of illegal distribution sites is expected to enable swift responses to copyright infringements.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2022.07a
• /
• pp.81-84
• /
• 2022

제조 산업에서의 이상치 검출은 생산품의 품질과 운영비용을 절감하기 위한 중요한 요소로 최근 딥러닝을 사용하여 자동화되고 있다. 이상치 검출을 위한 딥러닝 기법에는 CNN이 있으며, CNN을 계층적으로 구성할 경우 단일 CNN 모델에 비해 상대적으로 성능의 향상을 보일 수 있다는 것이 많은 선행 연구에서 나타났다. 이에 MVTec-AD 데이터셋을 이용하여 계층 CNN이 다중 클래스 이상치 판별 문제에 대해 효과적인지를 탐구하고자 하였다. 실험 결과 단일 CNN의 정확도는 0.7715, 계층 CNN의 정확도는 0.7838로 다중 클래스 이상치 판별 문제에 있어 계층 CNN 방식 접근이 다중 클래스 이상치 탐지 문제에서 알고리즘의 성능을 향상할 수 있음을 확인할 수 있었다. 계층 CNN은 모델과 파라미터의 개수와 리소스의 사용이 단일 CNN에 비하여 기하급수적으로 증가한다는 단점이 존재한다. 이에 계층 CNN의 장점을 유지하며 사용 리소스를 절약하고자 하였고 K-means, GMM, 계층적 클러스터링 알고리즘을 통해 제작한 새로운 클래스를 이용해 계층 CNN을 구성하여 각각 정확도 0.7930, 0.7891, 0.7936의 결과를 얻을 수 있었다. 이를 통해 Clustering 알고리즘을 사용하여 적절히 물체를 분류할 경우 물체에 따른 개별 상태 판단 모델을 제작하는 것과 비슷하거나 더 좋은 성능을 내며 리소스 사용을 줄일 수 있음을 확인할 수 있었다.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.10
• /
• pp.140-148
• /
• 1994

In this paepr, the modified neocognitron, which has capability of recognizing multi-patterns in input image in one pass, is proposed. It is the hierarchical neural network composed of S and V layer which are able to extract features and of C layer with little effect from deformation, changes in size, shifts in position. S and V cells extract the features of all patterns in input image by applying DCC(don't care condition) to those cells. S and C cells also have position informations of extracted patterns. Position information is used in extracting good features without extracted features beting interfered one another. The proposed method is shorter in recognition time than the selective attention method with backward connection, because of recognizing multi-patterns in one passe. The modified neocognitron can recognizze attached multi-patterns because of using DCC and position informations.

• Journal of Biomedical Engineering Research
• /
• v.27 no.5
• /
• pp.291-299
• /
• 2006

A Computer-Aided Diagnosis system has been examined to reduce the effort of radiologist. In this paper, we propose the algorithm using Support Vector Machine(SVM) classifier to discriminate whether microcalcifications are malignant or benign tumors. The proposed method to detect microcalcifications is composed of two detection steps each of which uses SVM classifier. The coarse detection step finds out pixels considered high contrasts comparing with neighboring pixels. Then, Region of Interest(ROI) is generated based on microcalcification characteristics. The fine detection step determines whether the found ROIs are microcalcifications or not by merging potential regions using obtained ROIs and SVM classifier. The proposed method is specified on Korean mammogram database. The experimental result of the proposed algorithm presents robustness in detecting microcalcifications than the previous method using Artificial Neural Network as classifier even when using small training data.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.9 no.7
• /
• pp.1588-1593
• /
• 2005

In coagulant control of water treatment plants, rule extraction, one of datamining categories, was performed for coagulant control of a water treatment plant. Clustering methods were applied to extract control rules from data. These control rules can be used for fully automation of water treatment plants instead of operator's knowledge for plant control. To perform fuzzy clustering, there are some coefficients to be determined and these kinds of studies have been performed over decades such as clustering indices. In this study, statistical indices were taken to calculate the number of clusters. Simultaneously, seed points were found out based on hierarchical clustering. These statistical approaches give information about features of clusters, so it can reduce computing cost and increase accuracy of clustering. The proposed algorithm can play an important role in datamining and knowledge discovery.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.15 no.4
• /
• pp.197-212
• /
• 2019

The purpose of this study is to identify the maturity stages of venture firms through classification analysis, which is widely used as a big data technique. Venture companies should develop a competitive advantage in the market. And the maturity stage of a company can be classified into five stages. I will analyze a difference in the growth stage of venture firms between the survey response and the statistical classification methods. The firm growth level distinguished five stages and was divided into the period of start-up and declines. A classification method of big data uses popularly k-mean cluster analysis, hierarchical cluster analysis, artificial neural network, and decision tree analysis. I used variables that asset increase, capital increase, sales increase, operating profit increase, R&D investment increase, operation period and retirement number. The research results, each big data analysis technique showed a large difference of samples sized in the group. In particular, the decision tree and neural networks' methods were classified as three groups rather than five groups. The groups size of all classification analysis was all different by the big data analysis methods. Furthermore, according to the variables' selection and the sample size may be dissimilar results. Also, each classed group showed a number of competitive differences. The research implication is that an analysts need to interpret statistics through management theory in order to interpret classification of big data results correctly. In addition, the choice of classification analysis should be determined by considering not only management theory but also practical experience. Finally, the growth of venture firms needs to be examined by time-series analysis and closely monitored by individual firms. And, future research will need to include significant variables of the company's maturity stages.

• Journal of Internet Computing and Services
• /
• v.20 no.6
• /
• pp.11-20
• /
• 2019

Numerous enterprises, organizations and individual users are exposed to large DDoS (Distributed Denial of Service) attacks. DDoS attacks are performed through a BotNet, which is composed of a number of computers infected with a malware, e.g., zombie PCs and a special computer that controls the zombie PCs within a hierarchical chain of a command system. In order to detect a malware, a malware detection software or a vaccine program must identify the malware signature through an in-depth analysis, and these signatures need to be updated in priori. This is time consuming and costly. In this paper, we propose a botnet detection scheme that does not require a periodic signature update using an artificial neural network model. The proposed scheme exploits Word2Vec and accelerated hierarchical density-based clustering. Botnet detection performance of the proposed method was evaluated using the CTU-13 dataset. The experimental result shows that the detection rate is 99.9%, which outperforms the conventional method.

• Journal of Internet Computing and Services
• /
• v.22 no.1
• /
• pp.23-30
• /
• 2021

This paper proposes a noise-tolerant image classification system using multiple autoencoders. The development of deep learning technology has dramatically improved the performance of image classifiers. However, if the images are contaminated by noise, the performance degrades rapidly. Noise added to the image is inevitably generated in the process of obtaining and transmitting the image. Therefore, in order to use the classifier in a real environment, we have to deal with the noise. On the other hand, the autoencoder is an artificial neural network model that is trained to have similar input and output values. If the input data is similar to the training data, the error between the input data and output data of the autoencoder will be small. However, if the input data is not similar to the training data, the error will be large. The proposed system uses the relationship between the input data and the output data of the autoencoder, and it has two phases to classify the images. In the first phase, the classes with the highest likelihood of classification are selected and subject to the procedure again in the second phase. For the performance analysis of the proposed system, classification accuracy was tested on a Gaussian noise-contaminated MNIST dataset. As a result of the experiment, it was confirmed that the proposed system in the noisy environment has higher accuracy than the CNN-based classification technique.

• Journal of Broadcast Engineering
• /
• v.26 no.4
• /
• pp.429-440
• /
• 2021

With the recent development of deep convolutional neural network learning, deep learning techniques applied to single image super-resolution are showing good results. One of the existing deep learning-based super-resolution techniques is RDN(Residual Dense Network), in which the initial feature information is transmitted to the last layer using residual dense blocks, and subsequent layers are restored using input information of previous layers. However, if all hierarchical features are connected and learned and a large number of residual dense blocks are stacked, despite good performance, a large number of parameters and huge computational load are needed, so it takes a lot of time to learn a network and a slow processing speed, and it is not applicable to a mobile system. In this paper, we use the residual dense structure, which is a continuous memory structure that reuses previous information, and the residual dense channel attention block using the channel attention method that determines the importance according to the feature map of the image. We propose a method that can increase the depth to obtain a large receptive field and maintain a concise model at the same time. As a result of the experiment, the proposed network obtained PSNR as low as 0.205dB on average at 4× magnification compared to RDN, but about 1.8 times faster processing speed, about 10 times less number of parameters and about 1.74 times less computation.