• Communications for Statistical Applications and Methods
• /
• v.28 no.1
• /
• pp.81-88
• /
• 2021

k-least quantile of squares (k-LQS) estimates are a generalization of least median of squares (LMS) estimates. They have not been used as much as LMS because their breakdown points become small as k increases. But if the size of outliers is assumed to be fixed LQS estimates yield a good fit to the majority of data and residuals calculated from LQS estimates can be a reliable tool to detect outliers. We propose to use LQS estimates for separating a clean set from the data in the context of outlyingness of the cases. Three procedures are suggested for the identification of outliers using LQS estimates. Examples are provided to illustrate the methods. A Monte Carlo study show that proposed methods are effective.

• International Journal of Internet, Broadcasting and Communication
• /
• v.14 no.2
• /
• pp.147-161
• /
• 2022

Video on demand (VOD) platforms provide immersive, inspiring, and commercial-free binge watching experiences. Recently, the number of these platform users increased dramatically as users can enjoy various contents without physical and time constraints during COVID-19. However, such platforms do not provide sufficient video censorship services while there is a strong need. In this study, we investigated the users' desire for video censorship when choosing and watching movies on VOD platforms, and how video censorship can be applied to different types of scenes to increase the censoring effect without diminishing the enjoyment. We first conducted an online survey with 98 respondents to identify the types of discomfort while watching sexual, violent, or drug-related scenes. We then conducted an in-depth online interview with 18 participants to identify the effective video filtering types and regions for each of the three scenes. Based on the findings, we suggest implications for designing a censor application for videos that contain uncomfortable scenes.

• Journal of Information Processing Systems
• /
• v.18 no.3
• /
• pp.344-358
• /
• 2022

Text summarization is the task of producing a shorter version of a long document while accurately preserving the main contents of the original text. Abstractive summarization generates novel words and phrases using a language generation method through text transformation and prior-embedded word information. However, newly coined words or out-of-vocabulary words decrease the performance of automatic summarization because they are not pre-trained in the machine learning process. In this study, we demonstrated an improvement in summarization quality through the contextualized embedding of BERT with out-of-vocabulary masking. In addition, explicitly providing precise pointing and an optional copy instruction along with BERT embedding, we achieved an increased accuracy than the baseline model. The recall-based word-generation metric ROUGE-1 score was 55.11 and the word-order-based ROUGE-L score was 39.65.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.17 no.4
• /
• pp.209-215
• /
• 2022

This paper proposes a novel approach for training an unsupervised depth estimation algorithm. The objective of unsupervised depth estimation is to estimate pixel-wise distances from camera without external supervision. While most previous works focus on model architectures, loss functions, and masking methods for considering dynamic objects, this paper focuses on the training framework to effectively use depth cue. The main loss function of unsupervised depth estimation algorithms is known as the photometric error. In this paper, we claim that direct depth cue is more effective than the photometric error. To obtain the direct depth cue, we adopt the technique of knowledge distillation which is a teacher-student learning framework. We train a teacher network based on a previous unsupervised method, and its depth predictions are utilized as pseudo labels. The pseudo labels are employed to train a student network. In experiments, our proposed algorithm shows a comparable performance with the state-of-the-art algorithm, and we demonstrate that our teacher-student framework is effective in the problem of unsupervised depth estimation.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2009.11a
• /
• pp.35-36
• /
• 2009

터치스크린 구동을 위해서는 터치된 위치의 아날로그 신호를 디지털 값으로 변환해 주는 ADC의 모드와 인터럽트 요청 발생을 위한 ADC의 채널에 연결된 스위치의 활성화 여부를 설정해야 한다. 본 논문은 터치스크린 LCD 모듈인 LP35가 부착된 LN2440SBC 임베디드 보드의 터치스크린 구동을 위한 초기화 방법과 터치스크린의 위치 좌표를 LCD의 픽셀 좌표로 변환해주는 좌표 보정 방법을 설명한다. 터치스크린의 구동은 크게 인터럽트 요청 대기/처리 상태로 구분된다. ADC 모드와 채널 스위치의 활성화 여부를, 인터럽트 요청 대기 상태에서는 인터럽트의 요청을 대기하며, 인터럽트 발생시 터치된 위치에 대한 아날로그 신호를 자동으로 디지털 값으로 변환하도록 설정하고, 인터럽트 요청 처리 상태에서는 인터럽트 요청을 처리하는 동안 새로운 인터럽트 요청을 마스킹(masking)하도록 설정한다. 좌표 보정은 터치된 위치 좌표 2개와 이에 대응되는 2개의 픽셀 좌표를 구하고, 이 좌표 값들을 이용하여 좌표 보정 일차 함수를 구함으로써 구현한다.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.461-463
• /
• 2003

The information on the cloud presence within a instantaneous field of view is the first step toward the derivation of many other geophysical parameters. Here, we first applied the current MODIS cloud detection algorithm developed by University of Wisconsin and compared the results to a visual interpretation of composite data, especially during the daytime. Most of cases, the detection algorithm performs very well, except a few cases with over-detection. One of the reasons for the false detection is due to the time independent use of land information which affects the threshold values of visible channel test. In the presentation, we show detailed analysis of the current cloud detection algorithm and suggest possible way to overcome the current shortfall.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2021.01a
• /
• pp.71-73
• /
• 2021

본 논문에서는 Scene Text Detection의 새로운 프레임워크인 TextFuseNet에 영상처리 관련 기술인 선명화 기법을 제안한다. Scene Text Detection은 야외 간판이나 표지판 등 불특정 배경에서 글자를 인식하는 기술이며, 그중 하나의 프레임워크가 TextFuseNet이다. TextFuseNet은 문자, 단어, 전역 기준으로 텍스트를 감지하는데, 여기서는 영상처리의 기술인 선명화 기법을 적용하여 TextFuseNet의 성능을 향상시키는 것이 목적이다. 선명화 기법은 기존 Sharpening Filter 방법과 Unsharp Masking 방법을 사용하였고 이 중 Sharpening Filter 방법을 적용하였을 때 AP가 0.9% 향상되었음을 확인하였다.

• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.509-513
• /
• 2021

개체명 인식이란, 문장에서 인명, 지명, 기관명, 날짜, 시간 등의 고유한 의미의 단어를 찾아서 미리 정의된 레이블로 부착하는 것이다. 일부 단어는 문맥에 따라서 인명 혹은 기관 등 다양한 개체명을 가질 수 있다. 이로 인해, 개체명에 대한 중의성을 가지고 있는 단어는 개체명 인식 성능에 영향을 준다. 본 논문에서는 개체명에 대한 중의성을 최소화하기 위해 사전을 구축하여 ELECTRA 기반 모델에 적용하는 학습 방법을 제안한다. 또한, 개체명 인식 데이터의 일반화를 개선시키기 위해 동적 마스킹을 이용한 데이터 증강 기법을 적용하여 실험하였다. 실험 결과, 사전 기반 모델에서 92.81 %로 성능을 보였고 데이터 증강 기법을 적용한 모델은 93.17 %로 높은 성능을 보였다. 사전 기반 모델에서 추가적으로 데이터 증강 기법을 적용한 모델은 92.97 %의 성능을 보였다.

• The Bulletin of The Korean Astronomical Society
• /
• v.45 no.1
• /
• pp.62.2-62.2
• /
• 2020

Globular clusters(GCs) are found not only around galaxies (galaxy GCs), but also between galaxies in galaxy clusters (intracluster GCs; ICGCs). The ICGCs, which are not bound to any of cluster member galaxies, are governed by the galaxy clutster potential. ICGCs have been detected in the wide field of Virgo and Fornax galaxy clusters. However, previous surveys covered only a small fraction of Coma and Perseus. In this study we present a wide field survey of these two galaxy clusters, using Subaru Hyper Suprime-Cam(HSC) archival images, covering a circular field with diameter of ~1.8 deg. We select ICGC candidates, by masking the images of bright galaxies and choosing point sources in the remaining area. We find thousands of ICGCs in each galaxy cluster. These ICGCs show a bimodal color distribution, which is dominated by blue GCs. We investigate spatial distributions and radial number density profiles of the blue and red ICGCs in each galaxy cluster. Implications of the results will be discussed.

• Journal of the Korean Society of Food Culture
• /
• v.22 no.2
• /
• pp.274-282
• /
• 2007

This study was performed to make new products rising Pimpinella brachycarpa N., one of the Korean aromatic medicinal plant called chamnamul, which is a perennial plant of the Umbelliferae family. New products were natural chamnamul spice, chamnamul soup, chamnamul tea, and chamnamul mook as functional foods. The masking effect of Pimpinella brachvcarpa N., on fishy and meaty odor were investigated to test the usefulness of chamnamul as a natural spice. It could be concluded that the effect of added amounts of chamnamul on the cream soup increases the taste and appearance, and improves the flavor and color by the sensory evaluation. The chamnamul tea prepared by a filtration method is better than that by a leaching method on the preference test. In the texture properties of chamnamul mooks by a texture analyzer (XT-RA, Texturometer), the cohessiveness of them was higher than that of the white one.