• 한국컴퓨터정보학회지
• /
• 제14권1호
• /
• pp.255-264
• /
• 2006

대부분의 비디오는 대용량의 장시간 데이터로서 비디오 시청자들이 전반적인 내용을 이해하기에는 충분하지 못하다. 본 논문에서는 이러한 문제점을 해결하기 위하여 효율적인 장면 전환 검출 방법과 새로운 샷 클러스터링을 이용한 비디오 개요 추출 방법을 제시한다. 장면전환 검출 방법은 컬러 히스토그램과 ${\chi}^2$ 히스토그램을 합성한 방법을 이용하여 추출하도록 한다. 클러스터링은 지역적 히스토그램의 차이값을 이용한 유사성 측정과 새로운 샷 병합 알고리즘을 통해 수행하도록 한다. 또한 실제 TV 방송 프로그램을 대상으로 비디오 개요 추출 실험 결과를 제시한다.

• 전기전자학회논문지
• /
• 제28권2호
• /
• pp.209-220
• /
• 2024

스케치는 직관적으로 정보를 표현할 수 있는 수단이나 실제 이미지와 비교하였을 때 추상화가 심하고 동일한 대상에 대한 스케치 임에도 변화가 매우 크다는 문제점이 있다. 따라서 스케치에 기반한 이미지 인식에 도움이 되는 특성을 표현하기 어려웠는데, 딥뉴럴 모델의 발전은 이미지와 스케치라는 두 도메인에 공통으로 존재하는 특성을 발견할 수 있게 해 주었다. 본 논문에서는 스케치에 기반한 이미지 검색 기술의 최신 연구 동향을 소개한다. 이미지 검색 외에 스케치에 기반한 이미지 인식과 이미지 생성의 연구 동향도 함께 요약하였으며, 스케치 기반 연구에서 주로 활용되는 데이터셋 및 해당 데이터셋에 기반하여 측정한 검색 성능을 포함하였다.

• 한국정밀공학회:학술대회논문집
• /
• 한국정밀공학회 1997년도 추계학술대회 논문집
• /
• pp.97-100
• /
• 1997

In order to predict the moldability of a injection molded part, a simulation of filling is needed. Especially when short shot is predicted by CAE simulation in the filling stage, there are mainly three ways to solve the problem. Modification of gate and runner, replacement of plastic resin, and adjustment of process conditions are the main ways. Among them, adjustment of process conditions is the most economic way in the cost and time since the mold doesn\\`t need t be modified at all. But it is difficult to adjust the process conditions appropriately in no times since it requires an empirical knowledge of injection molding. In this paper, a fuzzy neural network(FNN) based upon injection molding process is proposed to evaluate moldability in filling stage and also to solve the problem in case of short shot. An adequate mold temperature is generated through the fuzzy neural network where fill time and melt temperature are taken into considerations because process conditions affect each other.

• 한국컴퓨터정보학회논문지
• /
• 제7권4호
• /
• pp.75-82
• /
• 2002

대부분의 비디오는 대용량의 장시간 데이터로서 비디오 시청자들이 전반적인 내용을 이해하기에는 충분하지 못하다 본 논문에서는 이러한 문제점을 해결하기 위하여 효율적인 장면 전환 검출 방법과 새로운 샷 클러스터링을 이용한 비디오 개요 추출 방법을 제시한다. 장면전환 검출 방법은 컬러 히스토그램과 χ2 히스토그램을 합성한 방법을 이용하여 추출하도록 한다. 클러스터링은 지역적 히스토그램의 차이 값을 이용한 유사성 측정과 새로운 샷 병합 알고리즘을 통해 수행하도록 한다 또한 실제 TV방송 프로그램을 대상으로 비디오 개요 추출 실험 결과를 제시한다.

• 한국정보통신학회:학술대회논문집
• /
• 한국해양정보통신학회 2011년도 춘계학술대회
• /
• pp.599-601
• /
• 2011

POD (Print On Demand) 시스템은 출력 요청 시 즉시 출력할 수 있는 환경을 갖는 시스템을 의미한다. HTML 데이터를 이미지화하기 위해서는 COM 포트를 이용하여 수초의 시간이 필요한 문제가 발생한다. 하지만, 사용자가 감내할 만한 수준의 시간 지연을 통해 다양한 부가 서비스가 가능해진다. 본 논문에서는 웹 기반 화면 스크린 샷 메커니즘에 대해 기술하고, 이를 활용해 웹 사이트 미리보기 이미지 확보 방법과 POD시스템에서의 활용에 대해 기술한다.

• 한국정보디스플레이학회:학술대회논문집
• /
• 한국정보디스플레이학회 2006년도 6th International Meeting on Information Display
• /
• pp.1750-1755
• /
• 2006

Nonoptimal placement of short-channel-length TFTs in large-grained polycrystalline Si films with a periodic microstructure, as for instance obtained via 2-shot SLS, can potentially lead to degradation in the overall uniformity of the resultant devices. In this paper, we explain and demonstrate that by simply introducing a well-defined misorientation between the devices and the periodic microstructure, it is possible to significantly reduce (and potentially entirely eliminate) the device nonuniformity problem that can arise from such a cause.

• Design & Manufacturing
• /
• 제10권3호
• /
• pp.8-13
• /
• 2016

Despite technological advance, there have been several troubles in photo reaction injection molding (photo RIM) to produce ultra thin light guide panels (LGPs). In this study, micro leakage problem at cavity edge during photo RIM was investigated numerically. In order to obtain optimal processing conditions, we regulated inlet pressure of injected resin at the cavity edge and figured out micro leakage behaviors. At low inlet pressure (less than 100 Pa), though the micro leakage problem was not occurred, another problem, short shot due to not enough driving force, was appeared More than 1,000 Pa of the inlet pressure, injected resin was rapidly leaked through the micro gap at the cavity edge. Finally, we obtained optimal inlet pressure around 600 ~ 1,000 Pa. At this region, injected resin fully filled the cavity without micro leakage behavior. Based on the present study, further comparative investigations with experimental photo RIM should be performed to find optimal processing conditions for produce ultra thin LGPs.

• 한국정밀공학회:학술대회논문집
• /
• 한국정밀공학회 1995년도 춘계학술대회 논문집
• /
• pp.477-480
• /
• 1995

Dccisions of optimal filling conditions for the chip encapsulation have been done primarily by an ad hoc use of expertise accumulated over the years because the chip encapsulation process is quite complicated. The current CAE systems do not provide mold designers with necessary knowledge of the chip encapsulation for the successful design of optimal filling except flow simulation capability. There have been no attempts to solve the optimal filling problem in the process of the chip encapsulation. In this paper, we have constructed an design system for optimal filling to avoid short shot in the chip encapsulation process which combines an optimization methodology with CAE software.

• International journal of advanced smart convergence
• /
• 제8권4호
• /
• pp.68-74
• /
• 2019

Along with the deeper architecture in the deep learning approaches, the need for the data becomes very big. In the real problem, to get huge data in some disciplines is very costly. Therefore, learning on limited data in the recent years turns to be a very appealing area. Meta-learning offers a new perspective to learn a model with this limitation. A state-of-the-art model that is made using a meta-learning framework, Meta-SGD, is proposed with a key idea of learning a hyperparameter or a learning rate of the fast adaptation stage in the outer update. However, this learning rate usually is set to be very small. In consequence, the objective function of SGD will give a little improvement to our weight parameters. In other words, the prior is being a key value of getting a good adaptation. As a goal of meta-learning approaches, learning using a single gradient step in the inner update may lead to a bad performance. Especially if the prior that we use is far from the expected one, or it works in the opposite way that it is very effective to adapt the model. By this reason, we propose to add a weight term to decrease, or increase in some conditions, the effect of this prior. The experiment on few-shot learning shows that emphasizing or weakening the prior can give better performance than using its original value.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권10호
• /
• pp.3608-3626
• /
• 2021

The detection of bolts is an important task in high-speed train inspection systems, and it is frequently performed to ensure the safety of trains. The difficulty of the vision-based bolt inspection system lies in small sample defect detection, which makes the end-to-end network ineffective. In this paper, the problem is resolved in two stages, which includes the detection network and cascaded classification networks. For small bolt detection, all bolts including defective bolts and normal bolts are put together for conducting annotation training, a new loss function and a new boundingbox selection based on the smallest axis-aligned convex set are proposed. These allow YOLOv3 network to obtain the accurate position and bounding box of the various bolts. The average precision has been greatly improved on PASCAL VOC, MS COCO and actual data set. After that, the Siamese network is employed for estimating the status of the bolts. Using the convolutional Siamese network, we are able to get strong results on few-shot classification. Extensive experiments and comparisons on actual data set show that the system outperforms state-of-the-art algorithms in bolt inspection.