http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
- 中文 을 입력하시려면 zhongwen을 입력하시고 space를누르시면됩니다.
- 北京 을 입력하시려면 beijing을 입력하시고 space를 누르시면 됩니다.
RISS 인기검색어
- 검색키워드
- 저자 : 김재명(JaeMyung Kim) (검색결과 18 건)
- 무료
- 기관 내 무료
- 유료
-
-
한국개신교의 ‘생명평화’ 운동과 사상 -한국YMCA와 기독교환경운동연대를 중심으로-
김재명(Jaemyung Kim) 한국종교학회 2021 宗敎硏究 Vol.81 No.3
After democratization, Protestantism in Korea took the direction of a civic movement and its slogan was “Life-Peace.” The life-peace movement of Korean Protestants was led by ‘YMCA of KOREA’ and ‘Korea Christian Environmental Movement Solidarity for Integrity of Creation(KCEMS)’, but the activities and ideologies of the two groups are quite different. This means the differentiation of the Korean Protestant civic movement. Both flows are based on “Justice, Peace, and Integrity of Creation (JPIC),” but the way they understand the relationship among the three is different. The Life-Peace Movement of the YMCA Korea focuses on justice and peace rather than Integrity of Creation. On the other hand, KCEMS understands integrity of creation as the starting point and destination, and justice and peace in the context of integrity of creation. These differences are also revealed in the areas and methods of movement. YMCA Korea emphasizes justice and peace in the continuum of existing labor, human rights, economic, and social movements and additionally accepts ecological and environmental factors. On the other hand, KCEMS puts ecological and environmental factors at the center in their movement and accepts justice and peace in that context. The differences between the two groups are also revealed in the religious aspect. The Life-Peace Movement of the YMCA Korea started with a Christian ideology, but its direction is toward the so-called secular society or secular ethics. Therefore, in the process of developing a movement, they do not necessarily put Christianity in the forefront. On the other hand, the Life-Peace Movement of KCEMS takes the principle of ecology, combines it with the Christian ideology (creation-fall-salvation), and finds the principle of the movement in Christian thought (ecological spirituality, ecological theology). It moves toward spreading it as a movement within the church. This difference in ideological direction is also evident in the main movement areas of the two camps. While Korea YMCA mainly emphasizes social activities at the level of social movements, KCEMS as church movements covers not only the ecumenical camp but also the evangelical circle.
-
-
-
얼굴 감정인식을 위한 양자화된 경량 합성곱 신경망 구조 연구
김재명(Jaemyung Kim),강진구(Jin-Ku Kang),김용우(Yongwoo Kim) 대한전자공학회 2020 전자공학회논문지 Vol.57 No.12
최근 컴퓨터 비전 분야에서 우수한 성능을 보이는 CNN을 이용한 얼굴 감정인식 연구가 수행되고 있다. 높은 분류 정확도를 얻기 위해서는 많은 수의 파라미터와 높은 연산 복잡도를 가지는 CNN 구조가 필요하다. 하지만, 이와 같은 CNN 모델은 하드웨어 자원 사용량이 제한되어 있는 환경에서는 적합하지 않다. 따라서 본 논문에서는 제한된 환경 하에서의 최적화된 구현을 위해 적은 수의 파라미터와 낮은 연산 복잡도를 지닌 경량화된 CNN 구조를 설계하였고 정확도를 유지하면서도 파라미터 수 및 연산 복잡도를 낮출 수 있는 양자화 기법을 제안하였다. 또한 높은 분류 정확도를 위해 다양한 영상처리 알고리즘을 이용한 데이터 증강기법을 제안하였다. 제안한 부동소수점으로 훈련된 CNN모델(FP32)에 FERPlus 데이터 세트를 적용하여 성능을 평가한 결과, 파라미터 수는 약 1.98M개, FLOPs는 31MFLOPs, 정확도는 86.87%의 결과를 보였으며 기존의 경량화 모델과 비교하였을 때 가장 높은 정확도를 달성하였다. 또한, 제안한 양자화 기법을 적용하여 8-bit모델(INT8)에서 파라미터 수는 약 495K개, 4-bit모델(INT4)에서 파라미터 수는 약 247.5K개로 제안한 두 양자화된 CNN 모델(INT8, INT4)은 제안한 FP32 CNN모델 대비 1/4, 1/8만큼 적은 파라미터 수를 지니면서도 정확도 하락은 0.54% 이하인 것을 확인하였다. Recently, a study on facial emotion recognition using CNN, which has an excellent performance in the field of computer vision, is being conducted. To obtain high classification accuracy, a CNN structure with a large number of parameters and high computational complexity is required. However, such a CNN model is not suitable in an environment where the use of hardware resources is limited. In this paper, we designed a lightweight CNN structure with a small number of parameters and low computational complexity for an optimized implementation under a limited environment and proposed a quantization technique that can reduce the parameter size and computational complexity while maintaining accuracy. Also, for high classification accuracy, a data augmentation technique using various image processing algorithms was proposed. As a result of evaluating the performance by applying the FERPlus dataset to the proposed floating-point trained CNN model(FP32), the number of parameters was about 1.98M, FLOPs were about 31MFLOPs, and accuracy was about 86.87%. The highest accuracy was achieved compared to other lightweight models. In addition, two quantized CNN models(INT8, INT4) proposed by applying the proposed quantization technique as the number of parameters in the 8-bit model(INT8) are about 495K and the number of parameters in the 4-bit model(INT4) is about 247.5K. Compared to the proposed FP32 CNN model, it was confirmed that the number of parameters was as small as 1/4 and 1/8, but the accuracy drop was less than 0.54%.
-
김재명(Kim JaeMyung) 한국지역개발학회 1994 韓國地域開發學會誌 Vol.6 No.1
This paper investigates the feasibility of establishing an Inland Container Depot (ICD) in Chungcheong Region. The paper argues that Pusan Port, through which most of the containers exported from Korea, is highly congested. Therefore it costs unproportional transportstion expenses for Korean firms involved in exporting goods. In addition, the current transportation systems between the Capital Region and Pusan Port are ineffective. Since there are also increasing flows of materials within and nearby Chungcheong Region, it is necessary to establish and ICD in Chungcheong Region. This paper foresees that the location of an ICD in the Chungcheong Region creates addititional employment, income and service industries within Chungcheong Region, in addition to the reduction of transportation costs for the companies located in the Capital and Chungcheong Regions. For the specific location of the ICD, this research suggests Taejon and Chungjoo and concludes that both have pros and cons.
-
-
-
정수 연산만을 사용하는 하드웨어 친화적인 양자화된 CNN 구현
김재명(Jaemyung Kim),김용우(Yongwoo Kim) 대한전자공학회 2020 전자공학회논문지 Vol.57 No.12
최근 컴퓨터 비전 분야에서 CNN을 이용한 연구가 우수한 성능을 보여줌에 따라 보편적인 연구방법으로 자리 잡았다. 하지만 높은 성능을 얻기 위해서는 많은 수의 파라미터와 높은 연산 복잡도를 갖는 CNN 모델이 필요하다. 이러한 모델은 하드웨어 자원사용량에 제한이 있는 환경에서는 적합하지 않다. 따라서 CNN 모델의 구조를 유지한 채 정밀도를 낮추어 연산 복잡도 및 메모리사용량을 최적화할 수 있는 양자화 연구가 활발히 진행되고 있다. 하지만 대부분의 양자화 기법은 성능을 유지하기 위해 부동 소수점 스케일 인자를 사용하는 등 하드웨어 친화적이지 않은 방법들이 사용되었다. 따라서 본 논문에서는 하드웨어 친화적인 양자화 기법을 적용하여 양자화 인식 훈련을 수행하였고 양자화 인식 훈련으로 학습된 부동 소수점 파라미터를 정수 파라미터로 변환하였다. 그리고 오직 정수 연산만을 사용하는 기본 연산 블록들을 CNN으로 구성 및 계층 별 정밀도를 검증하는 기법을 제안한다. 8-bit로 양자화된 파라미터를 이용하여 정수 연산만을 사용하는 CNN의 성능을 CIFAR-10 데이터 세트로 평가한 결과, 부동 소수점 대비 파라미터 개수는 1/4로 줄었지만 정확도 하락은 0.71% 이하인 것을 확인하였다. 또한 기존의 양자화 인식 추론 기법과 비교하였을 때 추가적인 부동 소수점 연산기 대신 정수 연산기만을 사용하여 연산 복잡도를 낮출 수 있었으며 정확도 하락은 0.13% 이하인 것을 확인하였다. Recently, in the field of computer vision, research using CNN has established itself as a universal research method as it shows excellent performance. However, to obtain high performance, a CNN model with a large number of parameters and high computational complexity is required. This model is not suitable in an environment where hardware resource usage is limited. Therefore, while maintaining the structure of the CNN, quantization research that can optimize computational complexity and memory usage by lowering the precision is being actively conducted. However, most quantization techniques are not hardware-friendly methods were used, such as using a floating-point scale factor to maintain performance. In this paper, quantization aware training(QAT) was performed by applying a hardware-friendly quantization technique and floating-point parameters learned by QAT were converted into integer parameters. In addition, we proposed a method for constructing a CNN that combines basic operation blocks using integer arithmetic only operation and a method for verifying the precision of each layer. As a result of evaluating the performance of the proposed CNN using 8-bit quantized parameters with the CIFAR-10 dataset, it was confirmed that the number of parameters compared to floating-point was reduced to 1/4, but the accuracy drop was less than 0.71%. Besides, it was confirmed that the computational complexity could be reduced by using only an integer operator instead of an additional floating-point operator, and the accuracy drop was less than 0.13% when compared with the conventional quantization aware inference method.
-
DBpia연관 검색어 추천
이 검색어로 많이 본 자료
활용도 높은 자료
