• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송공학회 2010년도 추계학술대회
• /
• pp.78-81
• /
• 2010

In this paper, we propose an efficient and effective selective protection scheme to SVC that exploit the propagation of protection effect by protecting significant frames that can give the maximum visual quality degradation. We model SVC dependency coding structure as a directed acyclic graph which is characterized with an estimated visual quality value as the attribute at each node. The estimated visual quality is calculated by using our model based on the proportions of intra- and inter-predicted MBs, amounts of residual, and estimated visual quality of reference frames. The proposed selective protection scheme traverses the graph to find optimal protection paths that can give maximum visual quality degradation. Experimental results show that the proposed selective protection scheme reduces the required number of frames to be protected by 46.02% compared to the whole protection scheme and 27.56% compared to the layered protection scheme.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제14권9호
• /
• pp.3598-3614
• /
• 2020

With the increase of motor vehicles and tourism demand, some traffic problems gradually appear, such as traffic congestion, safety accidents and insufficient allocation of traffic resources. Facing these challenges, a model of Spatio-Temporal Dilated Convolutional Network (STDGCN) is proposed for assistance of extracting highly nonlinear and complex characteristics to accurately predict the future traffic flow. In particular, we model the traffic as undirected graphs, on which graph convolutions are built to extract spatial feature informations. Furthermore, a dilated convolution is deployed into graph convolution for capturing multi-scale contextual messages. The proposed STDGCN integrates the dilated convolution into the graph convolution, which realizes the extraction of the spatial and temporal characteristics of traffic flow data, as well as features of road occupancy. To observe the performance of the proposed model, we compare with it with four rivals. We also employ four indicators for evaluation. The experimental results show STDGCN's effectiveness. The prediction accuracy is improved by 17% in comparison with the traditional prediction methods on various real-world traffic datasets.

• 스마트미디어저널
• /
• 제12권1호
• /
• pp.9-16
• /
• 2023

최근 농가의 사과 품질 선별 작업에서 인적자원의 한계를 극복하기 위해 합성곱 신경망(CNN) 기반 시스템이 개발되고 있다. 그러나 합성곱 신경망은 동일한 크기의 이미지만을 입력받기 때문에 샘플링 등의 전처리 과정이 요구될 수 있으며, 과도 샘플링의 경우 화질 저하, 블러링 등 원본 이미지의 정보손실 문제가 발생한다. 본 논문에서는 위 문제를 최소화하기 위하여, 원본 이미지의 패치 기반 그래프를 생성하고 그래프 트랜스포머 모델의 랜덤워크 기반 위치 인코딩 방법을 제안한다. 위 방법은 랜덤워크 알고리즘 기반 위치정보가 없는 패치들의 위치 임베딩 정보를 지속적으로 학습하고, 기존 그래프 트랜스포머의 자가 주의집중 기법을 통해 유익한 노드정보들을 집계함으로써 최적의 그래프 구조를 찾는다. 따라서 무작위 노드 순서의 새로운 그래프 구조와 이미지의 객체 위치에 따른 임의의 그래프 구조에서도 강건한 성질을 가지며, 좋은 성능을 보여준다. 5가지 사과 품질 데이터셋으로 실험하였을 때, 다른 GNN 모델보다 최소 1.3%에서 최대 4.7%의 학습 정확도가 높았으며, ResNet18 모델의 23.52M보다 약 15% 적은 3.59M의 파라미터 수를 보유하여 연산량 절감에 따른 빠른 추론 속도를 보이며 그 효과를 증명한다.

• 한국안전학회지
• /
• 제31권5호
• /
• pp.141-148
• /
• 2016

There exists required safety integrity level (SIL) to assure safety in accordance with international standards for every electrical / electronics / control equipment or systems with safety related functions. The SIL is allocated from lowest level (level 0) to highest level (level 4). In order to guarantee certain safety level that is internationally acceptable, application of methodology for SIL allocation and demonstration based on related international standards is required. Especially, in case of the SIL allocation method without determining of quantitative tolerable risk, the additional review is needed to check whether it is suitable or not is required. In this study, the quantitative risk reduction model based on the safety integrity allocation results of railway platform screen door system using Risk Graph method has been examined in order to review the suitability of quantitative risk reduction according to allocated safety integrity level.

• 한국전자통신학회논문지
• /
• 제17권5호
• /
• pp.801-808
• /
• 2022

대규모 링크드 데이터에 어떻게 지식을 임베딩하고, 엔티티 매칭을 위해 어떻게 신경망 모델을 적용할 것인가에 대한 연구는 상대적으로 많이 부족한 상황이다. 이에 대한 가장 근본적인 문제는 서로 다른 레이블이 어휘 이질성을 초래한다는 것이다. 본 논문에서는 이러한 어휘 이질성 문제를 해결하기 위해 재정렬 구조를 결합한 확장된 GCN(Graph Convolutional Network) 모델을 제안한다. 제안된 모델은 기존 임베디드 기반 MTransE 및 BootEA 모델과 비교하여 각각 53% 및 40% 성능이 향상되었으며, GCN 기반 RDGCN 모델과 비교하여 성능이 5.1% 향상되었다.

• 한국컴퓨터정보학회논문지
• /
• 제24권7호
• /
• pp.61-69
• /
• 2019

In this paper, a classification method of random graph models is proposed and it is based on centralities of the random graphs. Similarity between two random graphs is measured for the classification of random graph models. The similarity between two random graph models $G^{R_1}$ and $G^{R_2}$ is defined by the distance of $G^{R_1}$ and $G^{R_2}$, where $G^{R_2}$ is a set of random graph $G^{R_2}=\{G_1^{R_2},...,G_p^{R_2}\}$ that have the same number of nodes and edges as random graph $G^{R_1}$. The distance($G^{R_1},G^{R_2}$) is obtained by comparing centralities of $G^{R_1}$ and $G^{R_2}$. Through the computational experiments, we show that it is possible to compare random graph models regardless of the number of vertices or edges of the random graphs. Also, it is possible to identify and classify the properties of the random graph models by measuring and comparing similarities between random graph models.

• ETRI Journal
• /
• 제39권2호
• /
• pp.255-263
• /
• 2017

This paper proposes a graph-theatrical approach to optimize spatial reuse by adopting a technique that quantizes the channel information into single bit sub-messages. First, we introduce an interference graph to model the network topology. Based on the interference graph, the computational requirements of the algorithm that computes the optimal spatial reuse factor of each user are reduced to quasilinear time complexity, ideal for practical implementation. We perform a resource allocation procedure that can maximize the efficiency of spatial reuse. The proposed spatial reuse scheme provides advantages in beamforming systems, where in the interference with neighbor nodes can be mitigated by using directional beams. Based on results of system level measurements performed to illustrate the physical interference from practical millimeter wave wireless links, we conclude that the potential of the proposed algorithm is both feasible and promising.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권8호
• /
• pp.3216-3230
• /
• 2015

A hot-issued research topic in the workflow intelligence arena is the emerging topic of "workflow-supported organizational social networks." These specialized social networks have been proposed to primarily represent the process-driven work-sharing and work-collaborating relationships among the workflow-performers fulfilling a series of workflow-related operations in a workflow-supported organization. We can discover those organizational social networks, and visualize its analysis results as organizational knowledge. In this paper, we are particularly interested in how to visualize the degrees of closeness centralities among workflow-performers by proposing a graphical representation schema based on the Graph Markup Language, which is named to ccWSSN-GraphML. Additionally, we expatiate on the functional expansion of the closeness centralization formulas so as for the visualization framework to handle a group of workflow procedures (or a workflow package) with organizational workflow-performers.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권12호
• /
• pp.3383-3397
• /
• 2023

Scene graphs serve as semantic abstractions of images and play a crucial role in enhancing visual comprehension and reasoning. However, the performance of Scene Graph Generation is often compromised when working with biased data in real-world situations. While many existing systems focus on a single stage of learning for both feature extraction and classification, some employ Class-Balancing strategies, such as Re-weighting, Data Resampling, and Transfer Learning from head to tail. In this paper, we propose a novel approach that decouples the feature extraction and classification phases of the scene graph generation process. For feature extraction, we leverage a transformer-based architecture and design an adaptive calibration function specifically for predicate classification. This function enables us to dynamically adjust the classification scores for each predicate category. Additionally, we introduce a Distribution Alignment technique that effectively balances the class distribution after the feature extraction phase reaches a stable state, thereby facilitating the retraining of the classification head. Importantly, our Distribution Alignment strategy is model-independent and does not require additional supervision, making it applicable to a wide range of SGG models. Using the scene graph diagnostic toolkit on Visual Genome and several popular models, we achieved significant improvements over the previous state-of-the-art methods with our model. Compared to the TDE model, our model improved mR@100 by 70.5% for PredCls, by 84.0% for SGCls, and by 97.6% for SGDet tasks.

• 인터넷정보학회논문지
• /
• 제23권5호
• /
• pp.145-154
• /
• 2022

자동 표적 인식(Automatic Target Recognition, ATR) 기술이 미래전투체계(Future Combat Systems, FCS)의 핵심 기술로 부상하고 있다. 그러나 정보통신(IT) 및 센싱 기술의 발전과 더불어 ATR에 관련이 있는 데이터는 휴민트(HUMINT·인적 정보) 및 시긴트(SIGINT·신호 정보)까지 확장되고 있음에도 불구하고, ATR 연구는 SAR 센서로부터 수집한 이미지, 즉 이민트(IMINT·영상 정보)에 대한 딥러닝 모델 연구가 주를 이룬다. 복잡하고 다변하는 전장 상황에서 이미지 데이터만으로는 높은 수준의 ATR의 정확성과 일반화 성능을 보장하기 어렵다. 본 논문에서는 이미지 및 텍스트 데이터를 동시에 활용할 수 있는 지식 그래프 기반의 ATR 방법을 제안한다. 지식 그래프와 딥러닝 모델 기반의 ATR 방법의 핵심은 ATR 이미지 및 텍스트를 각각의 데이터 특성에 맞게 그래프로 변환하고 이를 지식 그래프에 정렬하여 지식 그래프를 매개로 이질적인 ATR 데이터를 연결하는 것이다. ATR 이미지를 그래프로 변환하기 위해서, 사전 학습된 이미지 객체 인식 모델과 지식 그래프의 어휘를 활용하여 객체 태그를 노드로 구성된 객체-태그 그래프를 이미지로부터 생성한다. 반면, ATR 텍스트는 사전 학습된 언어 모델, TF-IDF, co-occurrence word 그래프 및 지식 그래프의 어휘를 활용하여 ATR에 중요한 핵심 어휘를 노드로 구성된 단어 그래프를 생성한다. 생성된 두 유형의 그래프는 엔터티 얼라이먼트 모델을 활용하여 지식 그래프와 연결됨으로 이미지 및 텍스트로부터의 ATR 수행을 완성한다. 제안된 방법의 우수성을 입증하기 위해 웹 문서로부터 227개의 문서와 dbpedia로부터 61,714개의 RDF 트리플을 수집하였고, 엔터티 얼라이먼트(혹은 정렬)의 accuracy, recall, 및 f1-score에 대한 비교실험을 수행하였다.