http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
백창환(Baek Chang Hwan),김연진(Kim Youn Jin),김홍석(Kim Hong-suk),박승옥(Park Seung-ok) 한국색채학회 2010 한국색채학회 논문집 Vol.24 No.3
인테리어 디자인산업에서 한쪽 벽면에 다른 벽들과는 다른 독특한 색이나 무늬를 넣는 포인트벽지가 각광을 받고 있다. 최근에는 LED 조명이 포인트벽지를 대체하기 시작하였다. LED 조명은 기존 조명등으로는 할 수 없었던 색상조절이 손쉬운 장점을 지녀 보다 인기를 끌고 있다. 본 연구에서는 LED 광색에 대한 색채감성을 두 가지 감성요소(‘동적인 - 정적인’과 ‘딱딱한 - 부드러운’) 로 분석하였다. 그 결과, LED 광색의 밝기가 높을수록 동적이고 딱딱한 감성에서 정적이고 부드러운 감성으로 바뀌는 것을 알 수 있었다. 이러한 경향은 모든 색상계열에서 나타났으며, 특히 빨간색상계열과 파란색상계열에서 두드러졌다. Interior design industry has focused on point wallpapering that decorates a side of the wall with distinct colors or patterns. Recently, light emitting diode (LED) lighting is often used to replace hardcopy point wallpapers. Color control of LED lighting is easier compared to regular bulbs. This study aims to analyze the LED color emotion in terms of two emotion factors; dynamic-static and hard-soft. As a result, LED color emotion is changed from dynamic and hard into static and soft when lightness increases. This tendency has been shown in all hue series especially for red and blue cases.
백창환(Chang-Hwan Baek),김연진(Youn-Jin Kim),김홍석(Hong-Suk Kim),박승옥(Seung-Ok Park) 한국조명·전기설비학회 2011 조명·전기설비학회논문지 Vol.25 No.1
Light emitting diode(LED) technology has been increasingly developed and larger color gamut by LED illuminations can be reproduced; therefore more efficient LED lighting design can be accomplished under a consideration of color emotion. Fifty-two LED colors which are uniformly distributed on the uniform chromaticity space are evaluated in terms of fatigueness and preference and their relation to three color-appearance attributes(lightness, chroma and hue) are investigated. As a result, 23 human observers likely to prefer and feel comfortable, when lightness of a given LED color stimulus increases as well as its chroma decreases. The highest fatigueness score is observed in red color series and the most preferred LED color is found in green color series. In addition, fatigueness and preference show a strong negative linear relation and their Pearson correlation is higher than -0.8.
Kobayashi 스케일과 I.R.I 스케일을 사용한 LED 광색의 형용사 이미지 분석
백창환(Chang-Hwan Baek),박승옥(Seung-Ok Park),김홍석(Hong-Suk Kim) 한국조명·전기설비학회 2011 조명·전기설비학회논문지 Vol.25 No.10
The aim of this study is to analyze the emotion adjectives for light emitting diode(LED) light colors using a twofold adjective image scales from Kobayashi and I.R.I. A set of psychophysical experiments using category judgment was conducted in an LED light color simulation system, in order to evaluate each emotion scale coordinate for those test light colors in both adjective image scales. In total, 49 test light colors from a combination of 6 color series were assessed by 15 human observers. As a result, Kobayashi adjective image scale clearly expressed to emotion adjectives of ‘Dynamic’, ‘Casual’, ‘Chic’, ‘Cool-casual’, ‘Modern’, and ‘Natural’ for different hues. In contrast, I.R.I adjective image scale expressed only 2 adjectives of ‘dynamic’ and ‘luxurious’ for the all hues.
한국인 데이터셋에 대한 얼굴 표정 감정 인식 모델 최적화
윤영철(Young-Chul Yoon),김산하(San-Ha Kim),김선호(Sun-Ho Kim),박희운(Hee-Woon Park),백창환(Chang-Hwan Baek) 대한전자공학회 2022 대한전자공학회 학술대회 Vol.2022 No.11
Recognizing facial emotion is an essential task to develop automatic tool such as service robot. A lot of methods are developed to recognize human emotion on multi-ethnic data sets. However, the previous methods show lower recognition performance when the models are applied on Korean dataset. To develop an automatic software for Korean, higher recognition ability on Korean emotion is required. Therefore, in this paper, we propose the optimized method to recognize Korean facial emotions using the Korean dataset. Through various experiments, we found that EfficientNetB4 with transfer learning and fine-tuning showed the best performance. The models were selected by considering GPU memory limitation. We compared this result with the model using both Korean datatset(AI Hub) and multi-ethnic dataset(AffectNet) in training and found the result that EfficientNetB4 trained by AI Hub showed the best performance. We named that model “Transferred fine-tuned EfficientNetB4 (TFE)”