• Journal of Korea Multimedia Society
• /
• v.22 no.9
• /
• pp.1000-1010
• /
• 2019

Image semantic segmentation and dehazing are key tasks in the computer vision. In recent years, researches in both tasks have achieved substantial improvements in performance with the development of Convolutional Neural Network (CNN). However, most of the previous works for semantic segmentation assume the images are captured in clear weather and show degraded performance under hazy images with low contrast and faded color. Meanwhile, dehazing aims to recover clear image given observed hazy image, which is an ill-posed problem and can be alleviated with additional information about the image. In this work, we propose a deep multi-task network for simultaneous semantic segmentation and dehazing. The proposed network takes single haze image as input and predicts dense semantic segmentation map and clear image. The visual information getting refined during the dehazing process can help the recognition task of semantic segmentation. On the other hand, semantic features obtained during the semantic segmentation process can provide cues for color priors for objects, which can help dehazing process. Experimental results demonstrate the effectiveness of the proposed multi-task approach, showing improved performance compared to the separate networks.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.6
• /
• pp.1778-1799
• /
• 2022

Automatic text summarization is a procedure that packs enormous content into a more limited book that incorporates significant data. Malayalam is one of the toughest languages utilized in certain areas of India, most normally in Kerala and in Lakshadweep. Natural language processing in the Malayalam language is relatively low due to the complexity of the language as well as the scarcity of available resources. In this paper, a way is proposed to deal with the text summarization process in Malayalam documents by training a model based on the Support Vector Machine classification algorithm. Different features of the text are taken into account for training the machine so that the system can output the most important data from the input text. The classifier can classify the most important, important, average, and least significant sentences into separate classes and based on this, the machine will be able to create a summary of the input document. The user can select a compression ratio so that the system will output that much fraction of the summary. The model performance is measured by using different genres of Malayalam documents as well as documents from the same domain. The model is evaluated by considering content evaluation measures precision, recall, F score, and relative utility. Obtained precision and recall value shows that the model is trustable and found to be more relevant compared to the other summarizers.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.32 no.2
• /
• pp.193-200
• /
• 2022

Unknown cryptographic communication protocols may have advantage of guaranteeing personal and data privacy, but when used for malicious purposes, it is almost impossible to identify and respond to using existing network security equipment. In particular, there is a limit to manually analyzing a huge amount of traffic in real time. Therefore, in this paper, we attempt to identify packets of unknown cryptographic communication protocols and separate fields comprising a packet by using machine learning techniques. Using sequential patterns analysis, hierarchical clustering, and Pearson's correlation coefficient, we found that the structure of packets can be automatically analyzed even for an unknown cryptographic communication protocol.

• Korean Journal of Radiology
• /
• v.22 no.3
• /
• pp.476-488
• /
• 2021

Objective: We aimed to develop a deep neural network for segmenting lung parenchyma with extensive pathological conditions on non-contrast chest computed tomography (CT) images. Materials and Methods: Thin-section non-contrast chest CT images from 203 patients (115 males, 88 females; age range, 31-89 years) between January 2017 and May 2017 were included in the study, of which 150 cases had extensive lung parenchymal disease involving more than 40% of the parenchymal area. Parenchymal diseases included interstitial lung disease (ILD), emphysema, nontuberculous mycobacterial lung disease, tuberculous destroyed lung, pneumonia, lung cancer, and other diseases. Five experienced radiologists manually drew the margin of the lungs, slice by slice, on CT images. The dataset used to develop the network consisted of 157 cases for training, 20 cases for development, and 26 cases for internal validation. Two-dimensional (2D) U-Net and three-dimensional (3D) U-Net models were used for the task. The network was trained to segment the lung parenchyma as a whole and segment the right and left lung separately. The University Hospitals of Geneva ILD dataset, which contained high-resolution CT images of ILD, was used for external validation. Results: The Dice similarity coefficients for internal validation were 99.6 ± 0.3% (2D U-Net whole lung model), 99.5 ± 0.3% (2D U-Net separate lung model), 99.4 ± 0.5% (3D U-Net whole lung model), and 99.4 ± 0.5% (3D U-Net separate lung model). The Dice similarity coefficients for the external validation dataset were 98.4 ± 1.0% (2D U-Net whole lung model) and 98.4 ± 1.0% (2D U-Net separate lung model). In 31 cases, where the extent of ILD was larger than 75% of the lung parenchymal area, the Dice similarity coefficients were 97.9 ± 1.3% (2D U-Net whole lung model) and 98.0 ± 1.2% (2D U-Net separate lung model). Conclusion: The deep neural network achieved excellent performance in automatically delineating the boundaries of lung parenchyma with extensive pathological conditions on non-contrast chest CT images.

• Cross-Cultural Studies
• /
• v.42
• /
• pp.261-283
• /
• 2016

The aim of this study is to consider the possibility of learning polite expressions of English using the mobile application Naver Band. To improve communicative competence, students need to develop sociolinguistic competence as well as grammatical and discourse competence. To be accordant to social context, the roles of the participants, and the purpose of the interaction, students have to make an appropriate utterance. When a hearer has a higher social status and is older than a speaker, or have low levels of familiarity, Korean native speaker tends to use polite expressions. Students need to learn polite expressions of English because English has a different honorific system from Korean. To realize the characteristics and function of polite language is one thing, but to learn it in a real classroom is another. This study attempts to apply the use of a mobile application, which is considered a user-friendly tool for students, into learning polite language using the UK historical drama, Downton Abbey. Two tests were administered to an experimental group that used the mobile application and a control group that used group work. The results of the two tests show that the use of the mobile application has a positive effect on learning polite expressions and is effective as an after-school activity. In an open-ended questionnaire, students tend to identify polite expressions as superior or high-level language forms and separate these forms from practical expressions. It should be noted as well that teachers need to consider instructing in modern English when using a historical drama in the teaching and learning of polite expressions.

• Journal of Korea Water Resources Association
• /
• v.56 no.9
• /
• pp.563-574
• /
• 2023

Recently, climate change has affected functional responses of river basins to meteorological variables, emphasizing the importance of rainfall-runoff simulation research. Simultaneously, the growing interest in machine learning has led to its increased application in hydrological studies. However, it is not yet clear whether machine learning models are more advantageous than the conventional conceptual models. In this study, we compared the performance of the conventional GR6J model with the machine learning-based Random Forest model across 38 basins in Korea using both gauged and ungauged basin prediction methods. For gauged basin predictions, each model was calibrated or trained using observed daily runoff data, and their performance was evaluted over a separate validation period. Subsequently, ungauged basin simulations were evaluated using proximity-based parameter regionalization with Leave-One-Out Cross-Validation (LOOCV). In gauged basins, the Random Forest consistently outperformed the GR6J, exhibiting superiority across basins regardless of whether they had strong or weak rainfall-runoff correlations. This suggest that the inherent data-driven training structures of machine learning models, in contrast to the conceptual models, offer distinct advantages in data-rich scenarios. However, the advantages of the machine-learning algorithm were not replicated in ungauged basin predictions, resulting in a lower performance than that of the GR6J. In conclusion, this study suggests that while the Random Forest model showed enhanced performance in trained locations, the existing GR6J model may be a better choice for prediction in ungagued basins.

• Korean Journal of Radiology
• /
• v.24 no.4
• /
• pp.294-304
• /
• 2023

Objective: We aimed to investigate whether image standardization using deep learning-based computed tomography (CT) image conversion would improve the performance of deep learning-based automated hepatic segmentation across various reconstruction methods. Materials and Methods: We collected contrast-enhanced dual-energy CT of the abdomen that was obtained using various reconstruction methods, including filtered back projection, iterative reconstruction, optimum contrast, and monoenergetic images with 40, 60, and 80 keV. A deep learning based image conversion algorithm was developed to standardize the CT images using 142 CT examinations (128 for training and 14 for tuning). A separate set of 43 CT examinations from 42 patients (mean age, 10.1 years) was used as the test data. A commercial software program (MEDIP PRO v2.0.0.0, MEDICALIP Co. Ltd.) based on 2D U-NET was used to create liver segmentation masks with liver volume. The original 80 keV images were used as the ground truth. We used the paired t-test to compare the segmentation performance in the Dice similarity coefficient (DSC) and difference ratio of the liver volume relative to the ground truth volume before and after image standardization. The concordance correlation coefficient (CCC) was used to assess the agreement between the segmented liver volume and ground-truth volume. Results: The original CT images showed variable and poor segmentation performances. The standardized images achieved significantly higher DSCs for liver segmentation than the original images (DSC [original, 5.40%-91.27%] vs. [standardized, 93.16%-96.74%], all P < 0.001). The difference ratio of liver volume also decreased significantly after image conversion (original, 9.84%-91.37% vs. standardized, 1.99%-4.41%). In all protocols, CCCs improved after image conversion (original, -0.006-0.964 vs. standardized, 0.990-0.998). Conclusion: Deep learning-based CT image standardization can improve the performance of automated hepatic segmentation using CT images reconstructed using various methods. Deep learning-based CT image conversion may have the potential to improve the generalizability of the segmentation network.

• Journal of Engineering Education Research
• /
• v.7 no.2
• /
• pp.40-50
• /
• 2004

To keep pace with rapidly changing social and technological trends, the Department of Mechanical and System Design Engineering at Hongik University is in the process of updating teaching materials and learning methodologies while evaluating/revising our existing curriculum. In order to be better apprised of the employment needs and student aspirations, we have conducted three separate but related surveys by targeting experienced professionals and managers, alumni working in industries as practising engineers, and our current students. The surveys have furnished us with valuable insights that, we hope, will lead to a new curriculum and learning methods more in tune with needs of our students as well as their future employers. Among other findings, the analysis of the survey responses points towards expansion of integrated laboratory and design experiences, cultivation of improved written and oral communication capabilities involving individual and team efforts, and development of interdisciplinary courses encompassing related fields of engineering as well as emerging fields such as nano and information technologies.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.30 no.4
• /
• pp.647-656
• /
• 2020

Recently, there has been an increasing number of cases in which important data (personal information, technology, etc.) of national and public institutions are leaked to the outside world. Surveys show that the largest cause of such leakage accidents is "insiders." Insiders of organization with the most authority can cause more damage than technology leaks caused by external attacks due to the organization. This is due to the characteristics of insiders who have relatively easy access to the organization's major assets. This study aims to present an optimized property selection model for detecting such abnormalities through supervised learning algorithms among machine learning techniques using actual data such as CrossNet data transfer system transmission log, e-mail transmission log, and personnel information, which safely transmits data between separate areas (security area and non-security area) of the business network and the Internet network.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.20 no.4
• /
• pp.826-832
• /
• 2016

In this paper, we propose a tongue diagnosis system for determining the presence of specific taste crack area as a first step in the digital tongue diagnosis system that anyone can use easily without special equipment and expensive digital tongue diagnosis equipment. Training DB was developed by the Haar-like feature, Adaboost learning on the basis of 261 pictures which was collected in Oriental medicine. Tongue candidate regions were detected from the input image by the learning results and calculated the average value of the HUE component to separate only the tongue area in the detected candidate regions. A tongue area is separated through the Connected Component Labeling from the contour of tongue detected. The palate regions were divided by the relative width and height of the tongue regions separated. Image on the taste area is converted to gray image and binarized with each of the average brightness values. A crack in the presence or absence was determined via Connected Component Labeling with binary images.