• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2019년도 춘계학술발표대회
• /
• pp.515-518
• /
• 2019

Semantic segmentation has remained as a challenging problem in the field of computer vision. Given the immense power of Convolution Neural Network (CNN) models, many complex problems have been solved in computer vision. Semantic segmentation is the challenge of classifying several pixels of an image into one category. With the help of convolution neural networks, we have witnessed prolific results over the time. We propose a convolutional neural network model which uses Fully CNN with deconvolutional pixel layers. The goal is to create a hierarchy of features while the fully convolutional model does the primary learning and later deconvolutional model visually segments the target image. The proposed approach creates a direct link among the several adjacent pixels in the resulting feature maps. It also preserves the spatial features such as corners and edges in images and hence adding more accuracy to the resulting outputs. We test our algorithm on Karlsruhe Institute of Technology and Toyota Technologies Institute (KITTI) street view data set. Our method achieves an mIoU accuracy of 92.04 %.

• 한국의류학회지
• /
• 제43권3호
• /
• pp.427-439
• /
• 2019

This study investigated the semantic network based on the focus of the fashion image and SNS text utilized by global SPA brands on the last seven years in terms of the quantity and quality of data generated by the fast-changing fashion trends and fashion content-based media environment. The research method relocated frequency, density and repetitive key words as well as visualized algorithms using the UCINET 6.347 program and the overall classification of the text related to fashion images on social networks used by global SPA brands. The conclusions of the study are as follows. A common aspect of global SPA brands is that by looking at the basis of text extraction on SNS, exposure through image of products is considered important for sales. The following is a discriminatory aspect of global SPA brands. First, ZARA consistently exposes marketing using a variety of professions and nationalities to SNS. Second, UNIQLO's correlation exposes its collaboration promotion to SNS while steadily exposing basic items. Third, in the case of H&M, some discriminatory results were found with other brands in connectivity with each cluster category that showed remarkably independent results.

• 대한건축학회연합논문집
• /
• 제21권5호
• /
• pp.109-116
• /
• 2019

The Seoul city recommended that old apartments be preserved, and as part of that, it decided to preserve some of the buildings for Jamsil Jugong, which was built in 1977, and Gaepo Jugong, which was constructed in 1981. The purpose of this study was to compare and review newspaper articles with two perspectives positive and negative about how the social perception of 'apartment trace remaining' was being constructed. By looking at the meaning of keywords delivered by newspaper articles and the interaction structure between keywords through the analysis of semantic networks, we analyzed how the media is pursuing an issue on the topic of preservation of architectural cultural heritage. The analysis results confirmed that there was a clear difference between positive and negative newspaper. Positive articles dealt with utilization from the point of view of keywords linked to preservation, and negative articles showed that keywords related to the property and backlash of residents linked to the policy of the Seoul Metropolitan Government were linked, leading to high negative public opinion.

• International Journal of Contents
• /
• 제17권1호
• /
• pp.1-10
• /
• 2021

Visual analytics is an emerging research field that combines the strength of electronic data processing and human intuition-based social background knowledge. This study demonstrates useful visual analytics with Tableau in conjunction with semantic network analysis using examples of sentiment flow and strategic communication strategies via Twitter in a blockchain domain. We comparatively investigated the sentiment flow over time and language usage patterns between companies with a good reputation and firms with a poor reputation. In addition, this study explored the relations between reputation and marketing communication strategies. We found that cryptocurrency firms more actively produced information when there was an increased public demand and increased transactions and when the coins' prices were high. Emotional language strategies on social media did not affect cryptocurrencies' reputations. The pattern in semantic representations of keywords was similar between companies with a good reputation and firms with a poor reputation. However, the reputable firms communicated on a wide range of topics and used more culturally focused strategies, and took more advantages of social media marketing by expanding their outreach to other social media networks. The visual big data analytics provides insights into business intelligence that helps informed policies.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권9호
• /
• pp.2458-2482
• /
• 2023

Medical image segmentation techniques based on convolution neural networks indulge in feature extraction triggering redundancy of parameters and unsatisfactory target localization, which outcomes in less accurate segmentation results to assist doctors in diagnosis. In this paper, we propose a multi-level semantic-rich encoding-decoding network, which consists of a Pooling-Conv-Former (PCFormer) module and a Cbam-Dilated-Transformer (CDT) module. In the PCFormer module, it is used to tackle the issue of parameter explosion in the conservative transformer and to compensate for the feature loss in the down-sampling process. In the CDT module, the Cbam attention module is adopted to highlight the feature regions by blending the intersection of attention mechanisms implicitly, and the Dilated convolution-Concat (DCC) module is designed as a parallel concatenation of multiple atrous convolution blocks to display the expanded perceptual field explicitly. In addition, MultiHead Attention-DwConv-Transformer (MDTransformer) module is utilized to evidently distinguish the target region from the background region. Extensive experiments on medical image segmentation from Glas, SIIM-ACR, ISIC and LGG demonstrated that our proposed network outperforms existing advanced methods in terms of both objective evaluation and subjective visual performance.

• 한국ITS학회 논문지
• /
• 제17권5호
• /
• pp.55-64
• /
• 2018

운전자의 안전사고에 직접적인 원인이 되고, 차량 파손을 유발시켜 재산상의 피해를 발생시키고 있는 포트홀을 완전 합성곱 신경망 기반의 자동으로 탐지하는 기법을 본 논문에서는 제안한다. 먼저, 실제 국내 도로를 주행하면서 차량에 설치된 카메라를 통하여 학습 데이터셋을 수집하고, 완전 합성곱 신경망 구조를 활용하여 의미론적 분할 형태로 신경망을 학습하였다. 어두운 환경에서 강건한 성능을 보이기 위하여 학습 데이터셋을 밝기에 따라서 증강하여 총 30,000장의 이미지를 학습하였다. 또한, 제안된 자동 포트홀 탐지 기술의 성능을 검증하기 위하여 총 450장의 평가 DB를 생성하였고, 총 네 명의 전문가가 각각의 이미지를 평가하였다. 평가 결과, 제안된 포트홀 탐지 기술은 높은 민감도 수치를 나타나는 것으로 평가 되었으며, 이는 정탐에서 강건한 성능을 보이는 것으로 해석 가능하다.

• 지식경영연구
• /
• 제19권3호
• /
• pp.47-61
• /
• 2018

A knowledge map, a taxonomy of knowledge repositories, must have capabilities supporting and enhancing knowledge user's activity to search and select proper knowledge for problem-solving. Conventional knowledge maps, however, have been hierarchically categorized, and could not support such activity that must coincide with the user's cognitive process for knowledge utilization. This paper, therefore, aims to verify and develop a methodology to build a networked knowledge map that can support user's activity to search and retrieve proper knowledge based on the referential navigation between content-relevant knowledge. This paper deploys keywords as the semantic information between knowledge, because they can represent the overall contents of a given document, and because they can play the role of semantic information on the link between related documents. By aggregating links between documents, a document network can be formulated: a keyword-based networked knowledge map can be finally built. Domain expert-based validation test was also conducted on a networked knowledge map of 50 research papers, which confirmed the performance of the proposed methodology to be outstanding with respect to the precision and recall.

• 한국통신학회논문지
• /
• 제19권6호
• /
• pp.1107-1114
• /
• 1994

본 논문은 적합성을 만족하는 프로토콜 변환 방법에 메시지간의 의미관계를 추가하여 프로토콜 변환기를 생성하는 방법을 제시한다.. 특히, 제안된 방법은 정형화된 기법과 하향식 방법을 도입하여 서로 다른 포로토콜로 구현된 컴퓨터 네트워크 사이의 프로토콜 불일치를 해결하는 효율적인 최다(maximal) 변환기를 유도한다. 따라서 제시된 방법을 사용하여 변환기를 유도할 경우, 불필요한 상태나 전이들의 상당 부분이 쉽게 제거된다.

• 한국정보처리학회논문지
• /
• 제7권4호
• /
• pp.1169-1174
• /
• 2000

Human-compute interface (HCl) should be designed with ease and flexibility for users for providing convenience and friendliness to users. The natural language processing is one of the aspects of HCl. This paper presents the method of design and implementation of SENKOV( Semantic Networks for Korean Verbs) System, which deals with isa hierarchies of Korean verbs for the natural language processing. The system's architecture is based on the differential theory and Levin verb classes. This paper selects about 600 Korean verbs which are commonly used in the daily life, and implements the SENKOV System. The experiments show that SENKOV has 44 top nodes and depth of about 2.35. In addition this paper applied the SENKOV System to co-occurrence constraint relationship among a adverbs and verbs, and proves the validity of the system. This paper is important in that it has made the first trial to classify Korean verb concept for HCl.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제5권11호
• /
• pp.549-554
• /
• 2016

컨셉넷은 일반상식을 노드(개념)와 에지(관계)로 표현해 놓은 그래프 형태의 지식 베이스이다. 완전한 지식 베이스를 구축하는 것은 매우 어려운 문제이기 때문에 지식 베이스는 미완결된 형태의 데이터를 담고 있는 경우가 많다. 불완전한 지식을 담고 있는 지식 베이스로부터의 추론 결과는 신뢰하기 어렵기 때문에 지식의 완결성을 높이기 위한 방법이 필요하다. 본 논문에서는 신경 텐서망을 이용하여 컨셉넷의 지식 미완결성 문제를 완화해 보고자 한다. 컨셉넷에서 추출한 사실주장(assertion)을 이용하여 신경 텐서망을 학습시킨다. 학습된 신경 텐서망은 두 개의 개념 정보를 입력으로 받고, 그 두 개념이 특정 관계로 연결될 수 있는지를 나타내는 점수값을 출력한다. 이와 같이 신경 텐서망은 노드들의 연결 차수(degree)를 높여, 컨셉넷의 완결성을 증대시킬 수 있다. 본 연구에서 학습시킨 신경 텐서망은 평가데이터에 대해서 약 87.7%의 정확도를 보였다. 또한 컨셉넷에 연결이 없는 노드 쌍에 대하여 85.01%의 정확도로 새로운 관계를 예측할 수 있었다.