• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제10권3호
• /
• pp.109-114
• /
• 2021

GAN(Generative Adversarial Network, 생성적 적대 신경망)은 이미지 생성모델로서 생성기 네트워크와 판별기 네트워크로 구성되며 실제 같은 이미지를 생성한다. GAN에 의해 생성된 이미지는 실제 이미지와 유사해야 하므로 생성된 이미지와 실제 이미지의 손실 오차를 최소화하는 손실함수(loss function)를 사용한다. 그러나 GAN의 손실함수는 이미지를 생성하는 학습을 불안정하게 만들어 이미지의 품질을 떨어뜨린다는 문제점이 있다. 이러한 문제를 해결하기 위해 본 논문에서는 GAN 관련 연구를 분석하고 에지 검출(edge detection)을 이용한 eGAN(edge GAN)을 제안한다. 실험 결과 eGAN 모델이 기존의 GAN 모델보다 성능이 개선되었다.

• 대한임베디드공학회논문지
• /
• 제17권2호
• /
• pp.77-83
• /
• 2022

In this paper, the two-stage object detection approach is proposed to implement a deep learning-based label inspection system on edge computing environments. Since the label printed on the products during the production process contains important information related to the product, it is significantly to check the label information is correct. The proposed system uses the lightweight deep learning model that able to employ in the low-performance edge computing devices, and the two-stage object detection approach is applied to compensate for the low accuracy relatively. The proposed Two-Stage object detection approach consists of two object detection networks, Label Area Detection Network and Character Detection Network. Label Area Detection Network finds the label area in the product image, and Character Detection Network detects the words in the label area. Using this approach, we can detect characters precise even with a lightweight deep learning models. The SF-YOLO model applied in the proposed system is the YOLO-based lightweight object detection network designed for edge computing devices. This model showed up to 2 times faster processing time and a considerable improvement in accuracy, compared to other YOLO-based lightweight models such as YOLOv3-tiny and YOLOv4-tiny. Also since the amount of computation is low, it can be easily applied in edge computing environments.

• 대한임베디드공학회논문지
• /
• 제15권6호
• /
• pp.299-305
• /
• 2020

We have developed a scheduler that additionally consider network performance by extending the Kubernetes developed to manage lots of containers in cloud computing nodes. The network delay adapt characteristics of the compute nodes were learned during server operation and the learned results were utilized to develop placement algorithm by considering the existing measurement units, CPU, memory, and volume together, and it was confirmed that the low delay network service was provided through placement algorithm.

• Genomics & Informatics
• /
• 제4권3호
• /
• pp.118-124
• /
• 2006

For the direct understanding of flow, pathway data are usually represented as directed graphs in biological journals and texts. Databases of metabolic pathways or signal transduction pathways inevitably contain these kinds of graphs to show the flow. KEGG, one of the representative pathway databases, uses the manually drawn figure which can not be easily maintained. Graph layout algorithms are applied for visualizing metabolic pathways in some databases, such as EcoCyc. Although these can express any changes of data in the real time, it exponentially increases the edge crossings according to the increase of nodes. For the understanding of genome scale flow of metabolism, it is very important to reduce the unnecessary edge crossings which exist in the automatic graph layout. We propose a metabolic pathway drawing algorithm for reducing the number of edge crossings by considering the fact that metabolic pathway graph is scale-free network. The experimental results show that the number of edge crossings is reduced about $37{\sim}40%$ by the consideration of scale-free network in contrast with non-considering scale-free network. And also we found that the increase of nodes do not always mean that there is an increase of edge crossings.

• Smart Structures and Systems
• /
• 제32권3호
• /
• pp.179-193
• /
• 2023

This study explores an alternative to the existing centralized process for data anomaly detection in modern Internet of Things (IoT)-based structural health monitoring (SHM) systems. An edge intelligence framework is proposed for the early detection and classification of various data anomalies facilitating quality enhancement of acquired data before transmitting to a central system. State-of-the-art deep neural network pruning techniques are investigated and compared aiming to significantly reduce the network size so that it can run efficiently on resource-constrained edge devices such as wireless smart sensors. Further, depthwise separable convolution (DSC) is invoked, the integration of which with advanced structural pruning methods exhibited superior compression capability. Last but not least, quantization-aware training (QAT) is adopted for faster processing and lower memory and power consumption. The proposed edge intelligence framework will eventually lead to reduced network overload and latency. This will enable intelligent self-adaptation strategies to be employed to timely deal with a faulty sensor, minimizing the wasteful use of power, memory, and other resources in wireless smart sensors, increasing efficiency, and reducing maintenance costs for modern smart SHM systems. This study presents a theoretical foundation for the proposed framework, the validation of which through actual field trials is a scope for future work.

• 정보처리학회논문지A
• /
• 제15A권6호
• /
• pp.345-350
• /
• 2008

최근에 병렬처리를 위한 새로운 위상으로 하이퍼 스타 연결망 HS(2n,n)가 제안되었다. 하이퍼 스타 연결망은 하이퍼큐브와 스타 그래프의 성질을 가지고 있으면서, 같은 노드수를 갖는 하이퍼큐브보다 망비용이 우수한 연결망이다. 본 논문에서는 하이퍼스타 연결망 HS(2n,n)에서 에지 중복 없는 스패닝 트리를 구성하는 알고리즘을 제안한다. 그리고 제안한 알고리즘에 의해 구성된 에지 중복 없는 스패닝 트리가 에지 중복없는 최적 스패닝 트리임을 증명한다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권9호
• /
• pp.4572-4586
• /
• 2019

Mobile Edge Computing (MEC) and Information-Centric Networking (ICN) are essential network architectures for the future Internet. The advantages of MEC and ICN such as computation and storage capabilities at the edge of the network, in-network caching and named-data communication paradigm can greatly improve the quality of video streaming applications. However, the packet loss in wireless network environments still affects the video streaming performance and the existing loss recovery approaches in ICN does not exploit the capabilities of MEC. This paper proposes a Deep Learning based Loss Recovery Mechanism (DL-LRM) for video streaming over MEC based ICN. Different with existing approaches, the Forward Error Correction (FEC) packets are generated at the edge of the network, which dramatically reduces the workload of core network and backhaul. By monitoring network states, our proposed DL-LRM controls the FEC request rate by deep reinforcement learning algorithm. Considering the characteristics of video streaming and MEC, in this paper we develop content caching detection and fast retransmission algorithm to effectively utilize resources of MEC. Experimental results demonstrate that the DL-LRM is able to adaptively adjust and control the FEC request rate and achieve better video quality than the existing approaches.

• 센서학회지
• /
• 제14권6호
• /
• pp.430-437
• /
• 2005

In this paper, we proposed a biologically inspired light-adaptive edge detection circuit based on the human retina. A saturating resistive network was suggested for light adaptation and simulated by using HSPICE. The light adaptation mechanism of the edge detection circuit was quantitatively analyzed by using a simple model of the saturating resistive element. A light-adaptive capability of the edge detection circuit was confirmed by using the one-dimensional array of the 128 pixels with various levels of input light intensity. Experimental data of the saturating resistive element was compared with the simulated results. The entire capability of the edge detection circuit, implemented with the saturating resistive network, was investigated through the two-dimensional array of the $64{\times}64$ pixels

• 한국정보과학회논문지:시스템및이론
• /
• 제36권5호
• /
• pp.429-436
• /
• 2009

오드 연결망은 그래프이론 모델의 하나로 발표되었는데, [1]에서 고장허용 다중컴퓨터에 대한 하나의 모형으로 소개되었고, 여러 가지 유용한 성질들 - 간단한 라우팅 알고리즘, 최대고장허용도, 노드 중복 없는 경로 등 - 이 분석되었다 본 논문에서는 오드 연결망 $O_d$ 에서 에지 중복 없는 스패닝 트리를 구성하는 알고리즘을 제안한다. 그리고 제안한 알고리즘에 의해 구성된 에지 중복 없는 스패닝 트리가 에지 중복 없는 최적 스패닝 트리임을 증명한다.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1998년도 추계학술대회 논문집 학회본부 B
• /
• pp.572-574
• /
• 1998

In this paper, we are proposed design of ATM networks to transfer for electric power system informations, proposed transport networks is partitioned management part and functional part, management part is partitioned edge network, core network, local network, authority network, functional part is partitioned core network, access network, edge area. It is based on laying and partitioning by ITU-T G.805, we also proposed ATM network requirements for Carrier Relay traffic acceptability in electric power system information.