DEA Window를 활용한 국내컨벤션센터의 효율성 변화 분석에 관한 연구

박진선 ( Jinsun Park ),김철원 ( Chulwon Kim ),김봉석 ( Bongseok Kim ) 한국질서경제학회 2015 질서경제저널 Vol.18 No.2

The purpose of this study is to analyze trends of six of Korea Convention Center’s efficiency of CCR, BCC models and scale efficiency through DEA Window from 2006 to 2013. In order to analyze the efficiency, I selected two inputs (input costs, tangible assets) and one output (revenue) then the average of efficiency trends and the stability of each convention center were examined. A research result shows that the efficiency of the technique in Korea Convention Centers is gradually increasing. The technical efficiency between the convention centers in Seoul and other provinces is experiencing a difference in efficiency. Convention centers in Seoul show a high efficiency and convention centers in province show a low efficiency. The convention industry is related not only to conferences but also to tourism, accommodation, transportation, catering and participation of residents. Thus, the convention centers in province seem to be needed a marketing strategy associated with the cultural heritage to hold conventions on the basis of infrastructure. A grade of pure technical efficiency is gradually declining contrary to the technical efficiency. In addition to the pure technical efficiency measurement, the scale efficiency is measured to examine the causes of the inefficiency of convention centers by the pure technical efficiency and the technical efficiency. The examination explains the result of inefficiency by the scale factor and the technique factor. As a result of the scale efficiency analysis, it is confirmed that the increasing trend of technical efficiency was affected by increasing the scale efficiency rather than the technique factor. Thus, this research suggests that convention centers should make an effort to improve the efficiency of technical factors based on the scale efficiency for the future. They should control investment of the excessive input cost and tangible assets. Through a marketing associated with a various convention sights, the revenue will be increased. Lastly, I hope this research helps improvement of Korea convention center management.

상용 수준의 자율주행 및 자동화를 위한 인공지능 기술 소개

박진선(Jinsun Park) 한국정보과학회 2022 정보과학회지 Vol.40 No.6

다양한 웹 브라우저 및 공간정보 오픈플랫폼 기반 동적정보 융합 성능 분석

박진선(Jinsun Park),위현중(Hyeonjoong Wi),강진원(Jinwon Kang),김민수(Minsoo Kim) 대한공간정보학회 2017 한국공간정보학회 학술대회 Vol.2017 No.10

개질형 수소제조장치의 소형화 및 고효율화

박강호(Kangho Park),박상혁(Sanghyuk Park),이임숙(Imsuk Lee),박진선(Jinsun Park) 한국가스학회 2021 한국가스학회 학술대회논문집 Vol.2021 No.11

Balanced L1 손실함수와 Brightness 손실함수에 따른 저조도 영상 개선 기법

박인철(Incheol Park),노정현(Jeonghyun Noh),박진선(Jinsun Park) 대한전자공학회 2023 대한전자공학회 학술대회 Vol.2023 No.6

In this paper, we propose balanced L1 loss and brightness loss to eliminate noise generated by the Zero-DCE. First, balanced L1 loss reduces image noise by preserving correlations between pixels. Brightness loss prevents the image from darkening when using balanced L1 loss. As a result of using two loss functions, NIQE was improved by 0.19 compared to the Zero-DCE, and the output image becomes brighter with less noise, making the image clearer.

사람 재식별을 위한 자기 유사성 특징 학습

권민성(Minseong Kweon),박진선(Jinsun Park) 대한전자공학회 2024 대한전자공학회 학술대회 Vol.2024 No.6

In this paper, we propose a novel person reidentification model named Self-Similarity Network (SSNet). Our method utilizes self-similarity features to learn the structural consistency between images of the same instance from different viewpoints. Specifically, this approach significantly improves feature learning for part-based features compared to traditional methods. To elaborate, before patch division, self-similarity is reflected to model the relationship between the upper and lower bodies. After patch division, structural features are enhanced by capturing self-similarity characteristics within each part of the body. These part-based features are used alongside global features and channel-based features to achieve robust person matching by accurately extracting features of the same person across different scenes. Our method demonstrated competitive performance in Market1501 and DukeMTMC benchmark datasets.

유압펌프의 오일 필터 막힘 고장진단에 대한 연구

최규정 ( Kyujeong Choi ),박진선 ( Jinsun Park ),조승제 ( Seungje Cho ),김정길 ( Jeonggil Kim ),김원균 ( Wongun Kim ),김영주 ( Youngjoo Kim ) 한국농업기계학회 2020 한국농업기계학회 학술발표논문집 Vol.25 No.1

유압시스템은 작동유를 압축하여 높은 압력을 만들어 큰 힘을 발생시킬 수 있으므로 농기계, 건설기계 등에서 널리 사용되고 있다. 작업 중에 유압시스템의 고장이 발생할 경우 작업이 중단되거나 더 큰 파손으로 이어져 경제적 손실이 발생할 수 있으므로 고장 발생을 초기에 진단하거나 발생 전에 예지하기 위한 기술의 필요성이 증가하고 있다. 본 연구에서는 건설기계인 드리프터의 고장 관련 데이터로부터 FMEA(Failure Modes and Effects Analysis)를 수행하여 주요고장부품으로 유압펌프를 선정하였으며 오일필터 막힘 현상을 고장모드로 선정하였다. 오일필터 막힘에 따른 상관관계를 확인하기 위하여 유압시스템 테스트벤치를 구성하였다. 펌프 흡입라인의 밸브를 잠가서 흡입필터 막힘을 모사하였으며 밸브를 40%~50% 닫았을 때 압력이 급격히 감소하는 현상을 확인하였다. 본 시험을 통하여 유압펌프 흡입라인의 압력센서를 사용하여 오일 필터 막힘 현상의 고장진단의 가능성을 확인하였으며 향후 실제 이물질 삽입을 통한 필터 막힘 구현시험을 진행할 예정이다.

뇌종양 분할을 위한 트랜스포머 기반의 다중 모달리티 지식 융합 방법

노정현(Jeonghyun Noh),박진선(Jinsun Park) 대한전자공학회 2024 대한전자공학회 학술대회 Vol.2024 No.6

Multi-modal magnetic resonance imaging (MRI) is widely employed for brain tumor diagnosis and analysis. However, conventional multi-modal brain tumor segmentation methods that leverage input fusion, where multiple modalities are input to the network simultaneously, face challenges in effectively exchanging information across modalities. In contrast, layer fusion methods, which take different modalities as input and exchange information within the network layers, facilitate cross-modality information exchange. Nevertheless, the concatenation operations commonly used in layer fusion inherently limit the ability to capture long-range dependencies. To address these issues, we propose a method that computes inter-modality differences to extract information to transfer across modalities and leverages transformers to facilitate this cross-modality information exchange. Experimental results on the BraTS2020 dataset show that ours achieved superior performance compared to previous methods.

혈관내피세포에서 c-Jun N-terminal kinase에 의해 조절되는 세포사멸에 고농도의 피노실빈이 미치는 효과

송지나(Jina Song),박진선(Jinsun Park),정은실(Eunsil Jeong),소아영(A-Young So),피재호(Jaeho Pyee),박헌용(Heonyong Park) 한국생명과학회 2015 생명과학회지 Vol.25 No.4

피노실빈은 소나무 등에서 흔히 관찰되는 스틸벤노이드이다. 혈관내피세포에서 ~pM에서 ~nM 정도의 낮은 농도의 피노실빈은 세포성장, 세포이동, 항염증반응을 유도한다. 그러나 최근에 고농도의 피노실빈이 소 대동맥 내피세포의 세포사를 유발하고 있음이 보고되었으나, 그 자세한 기전이나 경로 연구가 미흡하여 이번 연구에서 고농도 피노실빈의 세포사 유발 경로를 규명하기 위한 실험을 수행하였다. 고농도의 피노실빈은 caspase-3 절단, 포스파티딜 세린의 플립플롭, 핵 분절 등을 유도하는 것으로 보아 세포사멸을 통한 세포사를 유발함을 알았다. 또한 혈청기아나 100 μM etoposide에 의하여 촉발하는 caspase-3 활성/핵분절이 추가적으로 증가하는 현상이 관찰되었다. 이는 피노실빈이 일으키는 세포사멸은 혈청기아나 etoposide에 의해 일어나는 세포사멸과는 다른 경로로 진행됨을 의미하므로 고농도의 피노실빈에 의해 촉발되는 세포신호전달을 탐색한 결과, JNK와 eNOS가 관여함을 알았다. 두 신호전달 물질 중에 어느 것이 피노실빈 유도 세포사에 중요한 지를 알기 위한 추가적인 실험을 진행하였다. 그 결과, JNK 억제자인 SP-600125는 피노실빈에 의한 세포사멸을 억제하였으나, eNOS 억제자인 L-NAME은 아무런 영향이 없는 것으로 보아 JNK가 피노실빈에 의해 유발되는 세포사멸에 관여하는 세포신호 전달물질임을 알았다. Pinosylvin is a stilbenoid found in the Pinus species. Pinosylvin at ~pM to ~nM concentrations induces cell proliferation, cell migration and anti-inflammatory activity in endothelial cells. However, it was recently reported that pinosylvin at high concentrations (50 to 100 μM) induces cell death in bovine aortic endothelial cells. In this study, we conducted a series of experiments to discover how pinosylvin at a high concentration (50 μM) induces endothelial cell death. Pinosylvin at the high concentration was shown to induce endothelial cell apoptosis through enhancing caspase-3 activity, flip-flop of phosphatidyl serine, and nuclear fragmentation. We found that pinosylvin at the high concentration additively increased caspase-3 activity enhanced by serum-starvation or treatment with 100 μM etoposide. We also determined that pinosylvin at the high concentration promoted activations of c-Jun N-terminal kinase (JNK) and endothelial nitric oxide synthetase (eNOS). We further ran a series of experiments to find out which signaling molecule plays a critical role in the pinosylvin-induced apoptosis. We finally found that SP-600125, a JNK inhibitor, had an inhibitory effect on the pinosylvin-induced endothelial cell death, but L-NAME, an eNOS inhibitor, had no effect. These data indicate that JNK is involved in the pinosylvin-induced apoptosis. Collectively, pinosylvin at high doses induces cell apoptosis via JNK activation.

지식 증류를 이용한 Non-Local 기반의 가벼운 깊이 추정 네트워크

김장현(Janghyun Kim),박찬일(Chanill Park),박진선(Jinsun Park) 대한전자공학회 2024 대한전자공학회 학술대회 Vol.2024 No.6

Depth completion is a critical technique used to estimate dense depth information in surrounding environments for various applications such as autonomous driving and robotics. Real-world applications demand depth completion models with lightweight network architectures and reduced computational costs to ensure real-time operation. Therefore, we propose a non-local affinity distillation network for lightweight depth completion. Our approach emphasizes using a minimal number of influential neighbors with strong affinities to enable fast spatial propagation through optimal transport between teacher and student affinities via knowledge distillation. Experimental results demonstrate the effectiveness of our method, showcasing superiority over existing lightweight network while performing comparably to heavier depth completion methods.