• 전자통신동향분석
• /
• 제38권6호
• /
• pp.1-11
• /
• 2023

Large language models seem promising for handling reasoning problems, but their underlying solving mechanisms remain unclear. Large language models will establish a new paradigm in artificial intelligence and the society as a whole. However, a major challenge of large language models is the massive resources required for training and operation. To address this issue, researchers are actively exploring compact large language models that retain the capabilities of large language models while notably reducing the model size. These research efforts are mainly focused on improving pretraining, instruction tuning, and alignment. On the other hand, chain-of-thought prompting is a technique aimed at enhancing the reasoning ability of large language models. It provides an answer through a series of intermediate reasoning steps when given a problem. By guiding the model through a multistep problem-solving process, chain-of-thought prompting may improve the model reasoning skills. Mathematical reasoning, which is a fundamental aspect of human intelligence, has played a crucial role in advancing large language models toward human-level performance. As a result, mathematical reasoning is being widely explored in the context of large language models. This type of research extends to various domains such as geometry problem solving, tabular mathematical reasoning, visual question answering, and other areas.

• 한국정보통신학회:학술대회논문집
• /
• 한국해양정보통신학회 2005년도 추계종합학술대회
• /
• pp.595-598
• /
• 2005

MPEG-4 Visual은 객체 기반의 동영상 압축 국제 표준으로서 멀티미디어 통신에서부터 HDTV에 이르는 광범위한 응용 분야를 지원하기 위해 설계되었다. MPEG-4 표준안은 디코더에서 처리 가능한 디코딩 복잡도를 제한하기 위해 3가지 Video Buffering Verifier 모델을 정의하고 있다. 그 중 VCV 모델은 비트스트림을 매크로블록 단위로 디코딩하는 처리 속도에 대한 제한을 정의하고 있으며, 경계와 비경계 MB 두 가지만을 구별하는 VCV와 B-VCV 모델이 있다. 본 논문에서는 최적화된 MPEG-4 Reference Software를 이용하여 직사각형 객체와 임의 형상 객체 그리고 HDTV 해상도를 지원하는 다양한 코딩 타입에 대한 MB 디코딩 시간을 측정하여 디코딩 복잡도를 평가하였다. 실험결과 디코딩 복잡도가 코딩 타입에 따라 많은 차이가 있으며 디코더에서 이용 가능한 리소스의 더욱 효율적인 사용이 가능함을 보여주었다.

• 정보처리학회논문지B
• /
• 제18B권5호
• /
• pp.305-314
• /
• 2011

사람의 의도를 인지하기 위해 감정을 시각적으로 인식하는 연구는 전통적으로 감정을 드러내는 얼굴 표정을 인식하는 데 집중해 왔다. 최근에는 감정을 드러내는 신체 언어 즉 신체 행동과 자세를 통해 감정을 나타내는 방법에서 감정 인식의 새로운 가능성을 찾고 있다. 본 연구는 신경생리학의 시각계 처리 방법을 적용한 신경모델을 구축하여 행동에서 기본 감정 의도를 인식하는 방법을 제안한다. 이를 위해 시각 피질의 정보 처리 모델에 따라 생물학적 체계의 신경모델 검출기를 구축하여 신체 행동의 정적 자세에서 6가지 주요 기본 감정을 판별한다. 파라미터 변화에 강건한 제안 모델의 성능은 신체행동 자세 집합을 대상으로 사람 관측자와의 평가 결과를 비교 평가하여 가능성을 제시한다.

• 대한한의진단학회지
• /
• 제18권1호
• /
• pp.1-10
• /
• 2014

Objectives Visual inspection is the first diagnostic method in Oriental medicine, and visual inspection of eyes is the one among them. This study was written in order to complement further understanding on visual inspection of eyes. Methods 1. Out of 102 photographs submitted to the Society of HyungSang Medicine in 2009, 27 portrait pictures were selected as samples in blind by 2 clinicians. The samples were copied to make 54 sample pictures, and then randomly assigned to 4 clinicians. 2. The 4 clinicians evaluated the 54 samples for excess and deficiency of the eyes. The results were recorded as 5-points-scale, and their average and standard deviation was calculated. 3. Intra and inter class reliability test were measured using SPSS 13. Results For intra- and inter-class correlation coefficient (ICC) values were measured as 0.654~0.967 and 0.756~ 0.783 respectively, with the P-value below 0.05. Out of 27 originally selected samples, 7 pictures were selected as Deficiency Samples (with 3 pictures of male and 4 of females), and 20 as Excess Samples (with 4 of male and 16 of female). Among them, Sample No. 1, 9, 22, and 26 were selected as models of 'Excessive Eyes' for females, no. 4 and 5 as 'Very Excessive Eyes' for male and females, and no. 15 as 'Moderate Eyes' for females. Conclusion This study is the first attempt of quantitative measurement of excess and deficiency using the Visual Inspection of eyes by the visual inspection experts. Still, additional studies are needed regarding the relationship visual inspection methods have with existing standards of diagnosis.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1996년도 한국자동제어학술회의논문집(국내학술편); 포항공과대학교, 포항; 24-26 Oct. 1996
• /
• pp.89-92
• /
• 1996

In this paper, we developed visual tracking algorithm using double active bar. The active bar model to represent the object can reduce the search space of energy surface and better performance than those of snake model. However, the contour will not find global equilibrium when driving force caused by image may be weak. To overcome this problem. Double active bar is proposed for finding the global minimum point without any dependence on initialization. To achieve the goal, an deformable model with two initial contours in attempted to search for a global minimum within two specific initial contours. This approach improve the performance of finding the contour of target. To evaluate the performance, some experiments are executed. We can achieved the good result for tracking a object on noisy image.

• 한국가정과학회지
• /
• 제2권2호
• /
• pp.49-62
• /
• 1999

The purpose of this study is to evaluate the combination of details on the maternity clothes of one-piece type through visual evaluation which helps compensating pregnant women＇s body defects for their more attractive fashion styles. The data evaluated by a multiple ranking test were analyzed by mean, paired t-test, general linear models procedure and Duncan＇s multiple ranged test. The result are as follows : 1) The pregnant woman wearing the one-piece dress with notched collar, pleats and whole button looks longer in lower part of bodies, smaller in upper body, slimmer, have less appeared bust and abdomen, more balanced as a Whole than when wearing others. And also, it makes a pregnant woman be seen more refined and simple. 2) In case of a pregnant woman, a one-piece dress with notched collar, tuck and whole button makes her look longer in neck, narrower in shoulder, and more active than when wearing others. 3) With roll collar, pleats and whole button, it looks taller and more graceful than when wearing others.

• International Journal of Control, Automation, and Systems
• /
• 제6권4호
• /
• pp.607-612
• /
• 2008

This paper proposes a robot visual servo approach based on image appearance and a wavelet function neural network. The inputs of the wavelet neural network are changes of image features or the elements of image appearance vector, and the outputs are changes of robot joint angles. Image appearance vector is calculated by using eigen subspace transform algorithm. The proposed approach does not need a priori knowledge of the robot kinematics, hand-eye geometry and camera models. The experiment results on a real robot system show that the proposed method is practical and simple.

• 한국구조물진단유지관리공학회 논문집
• /
• 제2권3호
• /
• pp.235-244
• /
• 1998

In the design of bridges, the points of concem are the landscape design, the function, safety and economical efficiency. But most of studies have been performed on structural engineering. The study on the landscape design of bridges has not been done in korea. Therefore, in this research, the design method of bridges by the judgement of structural engineering and landscape engineering has been proposed, through the process to decide the shape of bridges. Also, the research studies a problem about the visual safety of the structural shape in the landscape design of bridges. The visual experiments applied to the seven models about the shape of hunch in bridge pier. The experiment was made in moving velocity of view point, steady looking time and track of eyeball movement.

• Fashion, Industry and Education
• /
• 제15권2호
• /
• pp.53-63
• /
• 2017

The growing significance of corporate social responsibility in the fashion industry has shed light on the importance of preparing fashion students to become socially responsible professionals. In spite of numerous benefits of service-learning, the teaching/learning method has been rarely employed in the fashion design and merchandising context. Therefore, the purpose of the study was first, to examine the concept and models of service-learning and compare different types of service-learning programs, and second, to discuss service-learning projects that were adopted in a visual design class as examples that service-learning can be effectively integrated into the fashion design and merchandising curriculum. This study provides the opportunity to share successful service-learning implementation with other educators to help with effective incorporation of the pedagogical program into the curriculum.

• 한국멀티미디어학회논문지
• /
• 제24권9호
• /
• pp.1279-1284
• /
• 2021

This paper addresses the task of transferring facial blendshape models to an arbitrary target face. Blendshape is a common method for the facial rig; however, production of blendshape rig is a time-consuming process in the current facial animation pipeline. We propose automatic blendshape facial rigging based on our blendshape transfer method. Our method computes the difference between source and target facial model and then transfers the source blendshape to the target face based on a deformation transfer algorithm. Our automatic method provides efficient production of a controllable digital human face; the results can be applied to various applications such as games, VR chating, and AI agent services.