• Journal of Information Processing Systems
• /
• 제19권1호
• /
• pp.98-108
• /
• 2023

This paper is an attempt to design segmentation method based on fully convolutional networks (FCN) and attention mechanism. The first five layers of the Visual Geometry Group (VGG) 16 network serve as the coding part in the semantic segmentation network structure with the convolutional layer used to replace pooling to reduce loss of image feature extraction information. The up-sampling and deconvolution unit of the FCN is then used as the decoding part in the semantic segmentation network. In the deconvolution process, the skip structure is used to fuse different levels of information and the attention mechanism is incorporated to reduce accuracy loss. Finally, the segmentation results are obtained through pixel layer classification. The results show that our method outperforms the comparison methods in mean pixel accuracy (MPA) and mean intersection over union (MIOU).

• Genomics & Informatics
• /
• 제12권1호
• /
• pp.35-41
• /
• 2014

Single-nucleotide polymorphisms (SNPs) have been emerging out of the efforts to research human diseases and ethnic disparities. A semantic network is needed for in-depth understanding of the impacts of SNPs, because phenotypes are modulated by complex networks, including biochemical and physiological pathways. We identified ethnicity-specific SNPs by eliminating overlapped SNPs from HapMap samples, and the ethnicity-specific SNPs were mapped to the UCSC RefGene lists. Ethnicity-specific genes were identified as follows: 22 genes in the USA (CEU) individuals, 25 genes in the Japanese (JPT) individuals, and 332 genes in the African (YRI) individuals. To analyze the biologically functional implications for ethnicity-specific SNPs, we focused on constructing a semantic network model. Entities for the network represented by "Gene," "Pathway," "Disease," "Chemical," "Drug," "ClinicalTrials," "SNP," and relationships between entity-entity were obtained through curation. Our semantic modeling for ethnicity-specific SNPs showed interesting results in the three categories, including three diseases ("AIDS-associated nephropathy," "Hypertension," and "Pelvic infection"), one drug ("Methylphenidate"), and five pathways ("Hemostasis," "Systemic lupus erythematosus," "Prostate cancer," "Hepatitis C virus," and "Rheumatoid arthritis"). We found ethnicity-specific genes using the semantic modeling, and the majority of our findings was consistent with the previous studies - that an understanding of genetic variability explained ethnicity-specific disparities.

• 한국문헌정보학회지
• /
• 제39권4호
• /
• pp.119-138
• /
• 2005

본 연구는 현행의 용어관계가 가진 문제점을 파악하기 위하여 용어관계의 다양한 사례를 조사 분석하고 이를 바탕으로 용어관계를 체계화하고자 하였다. 이를 위해 용어관계가 기반이 되는 분류, 시소러스, 주제명표목을 비롯하여 의미망, 온톨로지, 데이터베이스 등 기존의 여러 지식조직체계를 용어관계의 측면에서 재조명하여 그 특성 및 상호관계를 파악하였다. 또한 이들 지식조직체계에 실질적으로 나타나는 각종 용어관계의 사례와 용어관계에 대한 연구들을 광범위하게 수집하여 다양한 용어관계의 유형을 파악하였다. 이렇게 수집된 다양한 용어관계를 분석하여 실재하는 용어관계의 체계화 방안을 모색하였다.

• ETRI Journal
• /
• 제45권5호
• /
• pp.822-835
• /
• 2023

Recent semantic segmentation frameworks usually combine low-level and high-level context information to achieve improved performance. In addition, postlevel context information is also considered. In this study, we present a Context ReFinement Network (CRFNet) and its training method to improve the semantic predictions of segmentation models of the encoder-decoder structure. Our study is based on postprocessing, which directly considers the relationship between spatially neighboring pixels of a label map, such as Markov and conditional random fields. CRFNet comprises two modules: a refiner and a combiner that, respectively, refine the context information from the output features of the conventional semantic segmentation network model and combine the refined features with the intermediate features from the decoding process of the segmentation model to produce the final output. To train CRFNet to refine the semantic predictions more accurately, we proposed a sequential training scheme. Using various backbone networks (ENet, ERFNet, and HyperSeg), we extensively evaluated our model on three large-scale, real-world datasets to demonstrate the effectiveness of our approach.

• 한국건설관리학회논문집
• /
• 제23권3호
• /
• pp.45-55
• /
• 2022

BIM의 도입에 따라 공간이 개별 객체로 인식되면서 객체화된 공간의 속성정보는 법규검토, 에너지 분석, 피난 경로 분석 등을 위한 기반 데이터로 사용 가능하기에 BIM의 활용성을 넓힐 수 있는 발판을 마련하였다. 그러나 BIM 모델 내 개별 공간 속성의 오기입이나 누락이 없는 시멘틱 무결성(semantic integrity)이 보장되어야 하는데, 다수의 참여자에 의한 수작업으로 진행되는 BIM 모델링 과정 특성 상 설계 오류가 빈번히 발생한다는 문제점이 존재한다. 이를 해결하기 위해 BIM 모델의 공간 정합성 검증을 위한 연구가 다수 진행되었으나, 적용 범위가 한정적이거나 분류 정확도가 낮은 한계점이 존재하였다. 본 연구에서는 공간의 기하정보 뿐 아니라 BIM 모델 내 공간과 부재 간 연결 관계를 Graph Convolutional Networks (GCN) 학습과정에 활용하여 향상된 성능의 공간 자동 분류모델을 구축하고자 하였다. 구축된 GCN 기반 모델의 성능을 공간의 기하정보만으로 학습된 기계학습 모델인 Multi-Layer Perceptron (MLP)과 비교하여 공간 분류 시 연결 관계 적용의 효용성을 검증하고자 하였다. 이를 통해 관계정보 활용 시 약 8% 내외 수준으로 공간 분류 성능이 향상되는 것으로 확인되었다.

• 지능정보연구
• /
• 제14권4호
• /
• pp.19-30
• /
• 2008

온톨로지 기반 분산 정보시스템 환경에서는 시스템들 간의 자동화된 정보 공유를 지원하기 위해서 온톨로지 간의 의미적 이질성(semantic heterogeneity)을 해결해야 한다. 일반적으로 전문가들에게 미리 온톨로지들 간의 명시적 매핑(explicit mapping between ontologies)을 요청하고 있다. 하지만, 온톨로지 매핑의 고비용성 문제 때문에 모든 정보시스템들의 온톨로지 간의 일대일 매핑이 이루어지기 힘들다. 따라서, 본 논문에서는 분산된 온톨로지들 간의 매핑정보를 수집하고 재사용(reuse) 하고자 한다. 즉, 분산 환경에서 활용 가능한 온톨로지 매핑 정보의 결합(composition)을 위한 방법론을 제안한다. 이를 통해 주어진 두 온톨로지 간의 매핑 정보를 간접적으로 예측할 수 있게 된다. 특히, 본 연구에서 제안하는 온톨로지 매핑 결합 기법의 성능 평가를 위하여, 자동화된 쿼리 전송(propagation) 및 변환 (transformation) 시스템에 적용하였다.

• 국제학술발표논문집
• /
• The 9th International Conference on Construction Engineering and Project Management
• /
• pp.752-759
• /
• 2022

The instantiation of spaces as a discrete entity allows users to utilize BIM models in a wide range of analyses. However, in practice, their utility has been limited as spaces are erroneously entered due to human error and often omitted entirely. Recent studies attempted to automate space allocation using artificial intelligence approaches. However, there has been limited success as most studies focused solely on the use of geometric features to distinguish spaces. In this study, in addition to geometric features, semantic relations between spaces and elements were modeled and used to improve space classification in BIM models. Graph Convolutional Networks (GCN), a deep learning algorithm specifically tailored for learning in graphs, was deployed to classify spaces via a similarity graph that represents the relationships between spaces and their surrounding elements. Results confirmed that accuracy (ACC) was +0.08 higher than the baseline model in which only geometric information was used. Most notably, GCN was able to correctly distinguish spaces with no apparent difference in geometry by discriminating the specific elements that were provided by the similarity graph.

• Journal of Information Processing Systems
• /
• 제15권1호
• /
• pp.1-6
• /
• 2019

Recently, big data and artificial intelligence (AI) based on communication systems have become one of the hottest issues in the technology sector, and methods of analyzing big data using AI approaches are now considered essential. This paper presents diverse paradigms to subjects which deal with diverse research areas, such as image segmentation, fingerprint matching, human tracking techniques, malware distribution networks, methods of intrusion detection, digital image watermarking, wireless sensor networks, probabilistic neural networks, query processing of encrypted data, the semantic web, decision-making, software engineering, and so on.

• ETRI Journal
• /
• 제34권4호
• /
• pp.591-601
• /
• 2012

Unstructured peer-to-peer (p2p) networks usually employ flooding search algorithms to locate resources. However, these algorithms often require a large storage overhead or generate massive network traffic. To address this issue, previous researchers explored the possibility of building efficient p2p networks by clustering peers into communities based on their social relationships, creating social-like p2p networks. This study proposes a social relationship p2p network that uses a measure based on Hebbian theory to create a social relation weight. The contribution of the study is twofold. First, using the social relation weight, the query peer stores and searches for the appropriate response peers in social-like p2p networks. Second, this study designs a novel knowledge index mechanism that dynamically adapts social relationship p2p networks. The results show that the proposed social relationship p2p network improves search performance significantly, compared with existing approaches.

• Journal of Contemporary Eastern Asia
• /
• 제13권2호
• /
• pp.15-39
• /
• 2014

This research conducted a before/after naturalistic field experiment, with the early Arab Spring as the treatment. Compared to before the early Arab Spring, after the observation period the associations became stronger among the Web terms: 'Jihad, Sharia, innovation, democracy and civil society.' The Western concept of civil society transformed into a central Islamist ideological component. At another level, the inter-nation network based on Jihad-weighted Web hyperlinks between pairs of 46 Muslim Majority (MM) nations found Iran in one of the top two positions of flow betweenness centrality, a measure of network power, both before and after early Arab Spring. In contrast, Somalia, UAE, Egypt, Libya, and Sudan increased most in network flow betweenness centrality. The MM 'Jihad'-centric word co-occurrence network more than tripled in size, and the semantic structure more became entropic. This media "cloud" perhaps billowed as Islamist groups changed their material-level relationships and the corresponding media representations of Jihad among them changed after early Arab Spring. Future research could investigate various rival explanations for this naturalistic field experiment's findings.