• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권12호
• /
• pp.4439-4455
• /
• 2021

Compared with two-stage object detection algorithms, one-stage algorithms provide a better trade-off between real-time performance and accuracy. However, these methods treat the intermediate features equally, which lacks the flexibility to emphasize meaningful information for classification and location. Besides, they ignore the interaction of contextual information from different scales, which is important for medium and small objects detection. To tackle these problems, we propose an image pyramid network based on dual attention mechanism (DAIPNet), which builds an image pyramid to enrich the spatial information while emphasizing multi-scale informative features based on dual attention mechanisms for one-stage object detection. Our framework utilizes a pre-trained backbone as standard detection network, where the designed image pyramid network (IPN) is used as auxiliary network to provide complementary information. Here, the dual attention mechanism is composed of the adaptive feature fusion module (AFFM) and the progressive attention fusion module (PAFM). AFFM is designed to automatically pay attention to the feature maps with different importance from the backbone and auxiliary network, while PAFM is utilized to adaptively learn the channel attentive information in the context transfer process. Furthermore, in the IPN, we build an image pyramid to extract scale-wise features from downsampled images of different scales, where the features are further fused at different states to enrich scale-wise information and learn more comprehensive feature representations. Experimental results are shown on MS COCO dataset. Our proposed detector with a 300 × 300 input achieves superior performance of 32.6% mAP on the MS COCO test-dev compared with state-of-the-art methods.

• Journal of Korea Trade
• /
• 제25권2호
• /
• pp.75-94
• /
• 2021

Purpose - This article will take the Korean manufacturing industry as an example to estimate Korea's global value chain status from the perspective of overall and sub-industry, hoping to provide a theoretical reference for Korean manufacturing to climb the global value chain. Design/methodology - Based on the WIOD data. The data is calculated by using MATLAB (2014a) coding. The data for 6 sectors are classified according to the International Standard Industrial Classification revision 3 (ISIC Rev. 3), the WIOD data are used to calculate and compare the position, participation and dynamics of the Korea, China and USA' manufacturing industry in the 1995-2016. Findings - The empirical results supported conclusions of the theoretical model. In the Korean GVC of electrical and optical sector, while stronger forward linkages than backward linkages to GVC are advantageous for an average advanced country, the benefits of downstream tasks are pronounced for non-advanced countries. And proved the correlation for an index to capture a country's upstream position or downstream position, it makes sense to compare that Korea's exports of intermediates in the same sector that are used by China and USA. Originality/value - The first is to re-examine the characteristics of South Korea's participation in global value chains under a more systematic and accurate theoretical framework, which provides a new empirical reference for related research; the second is to content covers of the manufacturing 6 sectors, so as to more completely describe the characteristics of Korean manufacturing's participation in global value chains; The value of this paper is providing empirical evidence of the effect of Korea's the GVC of manufacturing sectors. In the GVC of 6 sectors, first three have a higher position in the value chain and are in the upper middle and upper reaches of the GVC. The latter two have a low GVC position index, which has become the main sector that pulls down the overall position of Korea's manufacturing industry.

• 중소기업연구
• /
• 제42권3호
• /
• pp.43-75
• /
• 2020

본 연구는 중소기업학회에서 편찬한 '중소기업연구'에 게재된 논문을 분석하여 사업실패의 연구동향을 알아보는데 있다. 이를 위하여, 첫째, 문헌고찰을 통해, 해외의 연구동향과 주요 연구 주제를 탐색하고, 본 연구를 위한 분석의 틀을 작성하였다. 둘째, 1979년부터 2019년에 이르기까지, 중소기업연구에 편찬된 총 1,060편의 논문 중, 실패와 관련된 16편을 선정하고 분석하였다. 세 번째, 중소기업연구 이외의 한국의 실패 연구동향을 알아보기 위해, 키워드 분석으로 24편을 추가로 선정하여 분석하였다. 본 논문에서는 실패 연구의 동향을 총 5가지 큰 주제로 구분하여 분석하였다. (1) 실패예측, (2) 실패 전·후 감정, (3) 감정 이외 실패 비용, (4) 실패 원인, (5) 재창업 결정 및 성공요인이다. 기존연구가 가지는 함의를 살펴봄으로써, 향후 실패분야의 연구방향을 제시하고 있다.

• 한국융합학회논문지
• /
• 제13권2호
• /
• pp.249-255
• /
• 2022

본 연구는 최근 가공 불량 예측 방법으로 주목받고 있는 머신러닝 기반의 모델을 이용하여 CNC 가공 불량 발생의 실시간 예측을 위한 분석 프레임워크를 제안하고, 해당 프레임워크에 기반하여 XGBoost, CatBoost, LightGBM, 랜덤 포레스트, Extra Trees, SVM, k-최근접 이웃, 로지스틱 회귀 모델을 CNC 설비에 기본 내장된 센서들로부터 추출된 데이터에 적용 및 분석하였다. 분석 결과 XGBoost, CatBoost, LightGBM 모델이 동일하게 가장 우수한 정확도, 정밀도, 재현율, F1 점수, AUC 값을 보였으며, 이 중 LightGBM 모델이 소요 실행 시간이 가장 짧은 것으로 나타났다. 이러한 짧은 소요 실행 시간은 실 시스템 구축 비용 절감, 빠른 불량 예측에 따른 CNC 장비 파손 확률 감소, 전체적인 CNC 활용률 증가 등의 실무적 장점을 가지므로 LightGBM 모델이 기본 센서들만 설치된 CNC 설비에 적용 시 가공 불량 예측에 가장 효과적으로 판단된다. 또한 소요 실행 시간 및 컴퓨팅 파워의 제약이 없는 상황에서는 LightGBM, Extra Trees, k-최근접 이웃, 로지스틱 회귀 모형으로 구성된 앙상블 모델을 적용할 경우 분류 성능이 최대화됨을 확인하였다.

• Smart Structures and Systems
• /
• 제29권1호
• /
• pp.181-193
• /
• 2022

Data-driven structural health monitoring (SHM) of civil infrastructure can be used to continuously assess the state of a structure, allowing preemptive safety measures to be carried out. Long-term monitoring of large-scale civil infrastructure often involves data-collection using a network of numerous sensors of various types. Malfunctioning sensors in the network are common, which can disrupt the condition assessment and even lead to false-negative indications of damage. The overwhelming size of the data collected renders manual approaches to ensure data quality intractable. The task of detecting and classifying an anomaly in the raw data is non-trivial. We propose an approach to automate this task, improving upon the previously developed technique of image-based pre-processing on one-dimensional (1D) data by enriching the features of the neural network input data with multiple channels. In particular, feature engineering is employed to convert the measured time histories into a 3-channel image comprised of (i) the time history, (ii) the spectrogram, and (iii) the probability density function representation of the signal. To demonstrate this approach, a CNN model is designed and trained on a dataset consisting of acceleration records of sensors installed on a long-span bridge, with the goal of fault detection and classification. The effect of imbalance in anomaly patterns observed is studied to better account for unseen test cases. The proposed framework achieves high overall accuracy and recall even when tested on an unseen dataset that is much larger than the samples used for training, offering a viable solution for implementation on full-scale structures where limited labeled-training data is available.

• PNF and Movement
• /
• 제20권3호
• /
• pp.359-370
• /
• 2022

Purpose: Physical therapists are required to properly choose the most appropriate treatment for each patient within the framework of the International Classification of Functioning, Disability, and Health (ICF model). The aims of this study were to determine whether neurological physical therapists in clinical settings in South Korea know about the ICF model and to investigate the current trends of outcome measures (OMs) used by them. Methods: Two hundred and one physical therapists who worked with patients with neurological disorders participated in this study. The survey was conducted via e-mail and asked about commonly used OMs and the considerations for selecting OMs. Results: All physical therapists involved in this study responded completely, and 45.8% of participants learned about the ICF model, while 37.3% understood the detailed information related to the ICF model. The rest of the participants did not know or just heard about the ICF model. The most frequently used tools at the body function/structure level were the Range of Motion (98%), Manual Muscle Test (97%), Berg Balance Scale (83.1%), and Modified Ashworth Scale (70.6%) when allowing repetition. At the activity level, the 10-meter walk test (71.1%), 6-minute walk test (54.2%), and Functional Ambulatory Category (43.3%) were used, while the Activity-Specific Balance Confidence Scale (23.9%) was used at the participation level. There was a positive relationship between the number of tools used and years of work, as well as the level of understanding of the ICF model. Conclusion: The results of this study suggest that it is necessary to learn the ICF model in a clinical setting. In addition, the medical system needs to be modified to encourage physical therapists in South Korea to use proper OMs within the ICF model.

• 시스템엔지니어링학술지
• /
• 제18권1호
• /
• pp.66-83
• /
• 2022

Artificial Intelligence (AI) is to assumed to be one of next generation technology which determine technological competitiveness and strategic advantage of a certain country. By using the patent data, this study aims to have a comparative analysis of the technological competitiveness of evolving artificial intelligence at different stages of development among the five largest intellectual property offices in the world (IP5). For the analysis data, all AI technology patent data from 1956 to 2019 were utilized according to the classification system presented in the "WIPO 2019 Technology Trend: Artificial Intelligence" report published by the World Intellectual Property Organization (WIPO) in 2019. The results shows that China has already surpassed the United States in terms of the number of patent applications in the field of artificial intelligence technology. However, in the domains of the United States, Europe, Japan, and Korea, the technology competitiveness of the United States is far ahead of China. Interestingly, the rate of increase of Korea's technology competitiveness is also very fast, and it has been shown that the technology strength is ahead of China in non-Chinese domains. The significance of this study can be found in the fact that the temporal and spatial change process of technological competitiveness of significant countries in the field of artificial intelligence technology artificial intelligence was viewed as a macro-framework using the technology index (TS) the differences were compared.

• Computers and Concrete
• /
• 제31권3호
• /
• pp.185-195
• /
• 2023

Antalya is located south part of minor Asia, one of the biggest cities in Türkiye. As a result of population growth and vast migration to Antalya, many parts of the city that were not suitable for construction due to its geological conditions have become urban areas, and most of these urban areas are full of poorly engineered buildings. Poor engineering has been combined with unplanned urbanization, that causes utter vulnerability to disasters in Antalya. When an earthquake-prone city, Antalya faces with an earthquake risk, fear arises in society. To overcome this problem, it has become necessary to investigate the building stock, expressed in hundreds of thousands, in a fast and reliable way and then perform an urban transformation to create the perception of structural safety. However, the excessive building stock, labor, and economic problems made the implementation stage challenging and revealed the necessity of finding alternative solutions in the field. The present study presents a novel approach for assessment and model based on a rapid visual inspection method to transform areas under earthquake risk in Türkiye. The approach aimed to rank the interventions for decision-making mechanisms by making comparisons in the scale hierarchy. In the present study, to investigate the proposed approach, over 26,000 buildings were examined in Antalya, which is the fifth largest city in Türkiye that has a population of over 2.5 Million. In the results of the study, the risk classification was defined in the framework of building, block, street, neighborhood, and district scales.

• 한국미생물·생명공학회지
• /
• 제51권1호
• /
• pp.109-123
• /
• 2023

Preterm birth (PTB) is defined as giving birth prior to the 37th week of pregnancy and is a major cause of infant mortality. Studies have indicated that the vaginal microbiota's composition and its dysbiosis, particularly during pregnancy, may play a major role in PTB. While previous research work concentrated on well-studied microorganisms such as Lactobacillus, Prevotella, Gardnerella, various other microbes, and their significance in the vaginal microbiota's stability remain unknown. Moreover, current studies have focused primarily on the relative abundances of the microbes found, without considering their interactions with other members of the vaginal microbiota. In this work, we developed a novel computational approach and performed taxonomic classification of vaginal microbiota samples stratified longitudinally (Term/PTB) to observe compositional disparities and find underexamined microbes that may be contributing to PTB. Furthermore, we carried out a correlational analysis to build a microbial co-interaction network and investigated the functional implications of the genes present in both Term and PTB samples. The co-occurrence network revealed that Lactobacillus acts in solidarity to maintain the stability of the vaginal microbiota and did not have strong co-interactions with any of the other microbes. Similarly, microbes with strong interactions with Atopobium, a well-known marker microbe of PTB, were also observed. Additionally, several genes such as PTXA, FANCM, GPX, and DUSP were found to be playing an important role in the occurrence of PTB. This study provides a novel conceptual framework revealing distinct vaginal microbiota signatures that could be potential therapeutic targets for the prevention of PTB.

• 대한조선학회논문집
• /
• 제60권2호
• /
• pp.95-109
• /
• 2023

In the 4th industrial revolution, changes in the technological paradigm have had a direct impact on the maintenance system of ships. The 2-stroke low speed engine system integrates with the core equipment required for propulsive power. The Condition Based Management (CBM) is defined as a technology that predictive maintenance methods in existing calender-based or running time based maintenance systems by monitoring the condition of machinery and diagnosis/prognosis failures. In this study, we have established a framework for CBM technology development on our own, and are engaged in engineering-based failure analysis, data development and management, data feature analysis and pre-processing, and verified the reliability of failure mode DB using LSTM algorithms. We developed various simulated failure mode scenarios for 2-stroke low speed engine and researched to produce data on onshore basis test_beds. The analysis and pre-processing of normal and abnormal status data acquired through failure mode simulation experiment used various Exploratory Data Analysis (EDA) techniques to feature extract not only data on the performance and efficiency of 2-stroke low speed engine but also key feature data using multivariate statistical analysis. In addition, by developing an LSTM classification algorithm, we tried to verify the reliability of various failure mode data with time-series characteristics.