http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Channel Attention과 그룹 컨볼루션을 이용한 효율적인 얼굴 감정인식 CNN
이명오,윤의녕,고승현,조근식 한국정보과학회 2019 정보과학회논문지 Vol.46 No.12
ecently, studies using the convolutional neural network have been actively conducted to recognize emotions from facial expressions. In this paper, we propose an efficient convolutional neural network that solves the model complexity problem of the deep convolutional neural network used to recognize the emotions in facial expression. To reduce the complexity of the model, we used group convolution, depth-wise separable convolution to reduce the number of parameters, and the computational cost. We also enhanced the reuse of featuRecently, studies using the convolutional neural network have been actively conducted to recognize emotions from facial expressions. In this paper, we propose an efficient convolutional neural network that solves the model complexity problem of the deep convolutional neural network used to recognize the emotions in facial expression. To reduce the complexity of the model, we used group convolution, depth-wise separable convolution to reduce the number of parameters, and the computational cost. We also enhanced the reuse of features and channel information by using Skip Connection for feature connection and Channel Attention. Our method achieved 70.32% and 85.23% accuracy on FER2013, RAF-single datasets with four times fewer parameters (0.39 Million, 0.41 Million) than the existing model.res and channel information by using Skip Connection for feature connection and Channel Attention. Our method achieved 70.32% and 85.23% accuracy on FER2013, RAF-single datasets with four times fewer parameters (0.39 Million, 0.41 Million) than the existing model. 최근 얼굴 표정에서 감정을 인식하기 위한 문제에서 컨볼루션 신경망을 이용한 연구가 활발히 진행되고 있다. 본 논문에서는 사람의 얼굴 표정에서 나타나는 감정을 인식하기 위해 사용하는 딥 컨볼루션 신경망의 모델 복잡도(Complexity) 문제점을 해결한 효율적인 컨볼루션 신경망을 제안한다. 본 논문에서는 모델의 복잡도를 줄이기 위해 그룹 컨볼루션, 깊이별 분리 컨볼루션을 사용하여 파라미터 수와 연산량을 감소시키고 특징 연결을 위한 Skip Connection과 Channel Attention을 사용하여 특징의 재사용성과 채널 정보를 강화하였다. 제안하는 모델의 학습 파라미터 개수는 0.39 M(Million), 0.41 M으로 기존 모델에 비해 4배 이상 적은 수의 파라미터를 사용하여 FER2013, RAF-single 데이터셋에서 각각 70.32%, 85.23%의 정확도를 달성하였다.
태권도 품새 지도자의 코칭유형과 품새 선수의 성취목표성향 및 운동지속의사의 관계
이명오(Lee, Myeong-O),최관용(Choi, Kwan-Yong) 한국무예학회 2021 무예연구 Vol.15 No.3
이 연구는 태권도 품새 지도자의 코칭유형과 선수의 성취목표성향 및 운동지속의사의 관계를 검증하는데 목적이 있다. 이러한 연구목적을 달성하기 위해 2020년 대한태권도협회에 등록된 태권도 품새 선수를 모집단으로 설정하여 214명의 선수에게 자료를 수집하였다. 자료처리는 SPSS 22 통계프로그램을 사용하였으며, 연구목적에 따라 빈도분석, 탐색적 요인분석 및 신뢰도 분석, 상관관계분석, 다중회귀분석을 사용하여 분석한 결과 다음과 같은 결론을 얻었다. 첫째, 태권도 품새 지도자의 코칭유형은 선수의 성취목표성향에 부분적으로 긍정적인 영향을 미쳤다. 둘째, 태권도 품새 지도자의 코칭유형의 설명지도와 관리지도만 운동지속의사에 긍정적인 영향을 미쳤다. 셋째, 태권도 품새 선수의 성취목표성향은 운동지속의사에 긍정적인 영향을 미쳤다. This study aims to verify the relationships of the coaching type of instructors of Taekwondo Poomsae with player’s achievement goal orientation and intention to continue the exercise. To achieve this research purpose, this study set the Taekwondo Poomsae players registered in Korea Taekwondo Association in 2020 as a population to collect data from 214 players. For data processing, SPSS 22 statistical software was employed, and as a result of analyses using frequency analysis, exploratory factor analysis and reliability analysis, correlation analysis, and multiple regression analysis, according to the research purpose, the following conclusions were drawn. First, the coaching type of instructors of Taekwondo Poomsae had a partially positive impact on the player’s achievement goal orientation. Second, of the coaching type of instructors of Taekwondo Poomsae, only explanation directions and administration directions had positive impacts on intention to continue the exercise. Third, the Taekwondo Poomsae player’s achievement goal orientation had a positive impact on the intention to continue the exercise.
이명오(Lee Myeong O),정국현(Jeong Kook Hyun) 세계태권도문화학회 2014 세계태권도문화학회지 Vol.- No.9
The purpose of this study is to clarify improvement methods of the referee policy for the enhancement of fairness in Taekwondo Poomsae competitions. In order to serve this purpose, qualitative research using open surveys was implemented. External, internal criticism from expert groups were carried out to secure validity, and categorization of the inductive analysis method was used to extract results. The results based on these processes are as follows. Above all, problems in external factors that hinder the fairness of the referee can be categorized as follows. First, a rearrangement in the period and procedure time of domestic Taekwondo Poomsae competitions is necessary. Second, geographic, environmental problems considering the venue of domestic Taekwondo Poomsae competitions must be resolved. Third, new advertisement strategies of domestic Taekwondo Poomsae competitions for the invigoration of Poomsae is necessary. Next, problems in internal factors that hinder the fairness of the referee can be categorized as follows. First, individual referees evaluate all scoring criteria, and this causes inefficiency. Second, scores lack meaning due to the high subjectivity of individual referees. Third, there are stereotypes against sports athletes. Fourth, the lack of continuous education programs of the judgement and scoring criteria; solutions to resolve these problems and enhance fairness are as follows. First, the tasks and activities of referees must be scheme-independent. Second, R&D in technical supplement methods, such as video deciphering must be proceeded. Third, referee courses must be expanded and programs must be improved, reinforcing education purposes and objectives. Fourth, referees should be designated for respective scoring criteria. Fifth, a renewed Poomsae, breaking away from the official Poomsae must be introduced. There are high hopes that the measures to enhance the fairness of referees above may reinforce the fruitfulness and quality of the Poomsae field, which is contributing to the invigoration of Taekwondo, breaking away from the conventionally prominent competitions of domestic and international Taekwondo.
이명오 ( Lee Myeong-o ) 한국언어문화교육학회 2006 언어와 문화 Vol.2 No.3
As the number of female international marriage migrants increases, the Ministry of Gender Equality and Family has published a korean text book for beginners composed of female international marriage migrants. The text book has made to develop their korean skill and help them to adapt to korean society. Keimyung University conducted education with the text book published by Ministry of Gender Equality by Daegu Gieongbuk Development Institute. The paper explored insufficient facts and practical application of the text book through teaching at school. Moreover, this paper suggests the way how the text book should be improved in practical teaching. (Keimyung University)
고속도로 교통안전 AI 데이터셋을 활용한 교통흐름 분석 모델 개발
한서우(Han Seo Woo),이명오(Lee Myeong Oh),김용현(Kim Young Hyun),장수현(Jang Soo Hyun) 한국통신학회 2021 한국통신학회 학술대회논문집 Vol.2021 No.2
본 논문은 자동차 등록대수가 증가함에 따라 고속도로 내 사고를 줄이고, 원활한 차량흐름을 만들기 위해 CCTV 영상 기반 실시간 교통정보를 제공하는 방법을 연구하였다. 현재 해외 고속(일반)도로 CCTV 기반 객체검출, 객체 분할 데이터셋이 많지만, 국내 고속(일반)도로 환경의 CCTV 데이터셋은 거의 없다. 따라서 본 논문은 국내 고속(일반)도로 환경 기반 객체검출, 객체 분할을 인공지능 학습을 위한 고속도로 교통안전 AI 데이터셋 및 라이다와 영상을 이용한 VDL 방식 속도추정 ground truth(GT) 데이터셋을 제작하였다. 이 데이터셋을 이용하여 실시간 고속도로 교통흐름 분석 모델 개발하였으며 해당 모델의 출력값으로 실시간 교통분석을 위한 교통파라미터(차로별 평균속도, 차량별 속도, 차로별 교통량 등)를 얻어냈다.
시계열 교통데이터 기반 고속도로 교통흐름 예측 모델 연구
강동묵(DongMug Kang),이명오(MyungOh Lee),김용현(YongHyun Kim),윤상훈(SangHun Yoon),신대교(DaeKyo Shin),장수현(SooHyun Jang) 대한전자공학회 2021 대한전자공학회 학술대회 Vol.2021 No.6
There are short sections where congestion often occurs on the highway, such as the exit route of the highway. If traffic flow can be predicted in the future point for these sections, drivers can drive actively and efficiently in consideration of road conditions. In this paper, we propose study on traffic flow prediction based on time series data of highway and average speed is used as traffic flow. In the case of time series data, it is extracted through speed estimation and object detection algorithms from 10 CCTV video installed on Yeongdong highway(Maseong IC ~ Singal JC, 3.4km) in Korea. For traffic flow prediction, we use Conv2D-LSTM model that consider spatio-temporal features. Also, in order to intuitively and efficiently represent the traffic congestion degree of a highway, traffic congestion parameter is proposed. As a result, the prediction model shows performance with an error of 7.08 based on the MAE(Mean Absolute Error), and the speed at the future point can be predicted.