• Annual Conference on Human and Language Technology
• /
• 2019.10a
• /
• pp.424-426
• /
• 2019

정답 후보군 탐지 모델은 최근 활발히 진행되고 있는 질의-응답 데이터 수집 연구의 선행이 되는 연구로 특정 질문에 대한 정답을 주어진 단락에서 추출하는 작업을 말한다. 제안 모델은 포인터 네트워크 디코더를 통하여 기존의 순차 레이블링 모델에서 처리할 수 없었던 정답이 겹치는 문제에 대해서 해결할 수 있게 되었다. 그리고 독립된 두 개의 포인터 네트워크 디코더를 사용함으로써, 단일 포인터 네트워크로 처리할 수 없었던 정답의 탐지가 가능하게 되었다.

• KIISE Transactions on Computing Practices
• /
• v.23 no.9
• /
• pp.542-549
• /
• 2017

Sequence-to-sequence models and similar pointer networks suffer from performance degradation when an input is composed of multiple sentences or when the length of the input sentence is long. To solve this problem, this paper proposes a hierarchical pointer network model that uses both the word level and sentence level information to encode input sequences composed of several sentences at the word level and sentence level. We propose a hierarchical pointer network based coreference resolution that performs a coreference resolution for all mentions. The experimental results show that the proposed model has a precision of 87.07%, recall of 65.39% and CoNLL F1 74.61%, which is an improvement of 21.83% compared to an existing rule-based model.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2010.11a
• /
• pp.617-620
• /
• 2010

최근 고성능 모바일 단말기와 다양한 컨텐츠가 등장하면서 유비쿼터스 환경에서의 사용자 인터페이스에 대한 관심이 높아지고 있다. 특히 모바일 프로젝터는 장소의 제한을 받지 않고 큰 화면을 다른 사람과 공유할 수 있는 장점이 있는 반면에 단말기를 직접 제어해야 하는 불편함이 있다. 본 논문에서는 모바일 환경에서 카메라를 통해 입력된 영상 정보만으로 사용자가 스크린에 비추는 레이저 포인터를 실시간으로 검출하는 알고리즘을 제안한다. 제안하는 알고리즘은 색상 감지와 움직임 감지로 나뉜다. 단일 프레임에서 영상 성분의 평균을 이용한 조건을 검사하여 레이저 포인터 색상 영역을 검출하고, 인접한 프레임과 현재 프레임과의 차를 구하며 그 차이가 임계값보다 큰 영역을 움직임 영역으로 검출한다. 마지막으로 색상 검출 영역과 움직임 검출 영역을 동시에 만족하는 영역을 최종적으로 레이저 포인터 영역으로 인식한다. 본 기법은 영상 정보만 사용하기 때문에 센서나 불필요한 장비를 착용할 필요가 없고 영상 성분 평균을 이용하므로 프로젝터 성능에 따른 조도의 변화에 강건하여 효과적인 레이저 포인터 검출이 가능하다. 실험결과는 주변 조명의 밝기에 따라 차이가 있지만 대부분 80% 이상의 검출률과 16% 미만의 오검출률의 성능으로 나타났고, 이 같은 결과는 사용자의 주관적인 만족을 보장하였다.

• Journal of Digital Convergence
• /
• v.10 no.2
• /
• pp.211-216
• /
• 2012

There are two kinks of processes could detect the laser pointer. One is the process which detects the location of the pointer. the other one is a possibility of dividing with the process which converts the coordinate of the laser pointer which is input in coordinate of the monitor. The previous Mean-Shift algorithm is not appropriately for real-time video image to calculate many quantity. In this paper, we proposed the context driven real-time laser pointer detection and tracking. The proposed method is a possibility of getting the result which is fixed from the situation which the background and the background which are complicated dynamically move. In the actual environment, we can get to give constant results when the object come in, when going out at forecast boundary. Ultimately, this paper suggests empirical application to verify the adequacy and the validity with the proposed method. Accordingly, the accuracy and the quality of image recognition will be improved the surveillance system.

• KIISE Transactions on Computing Practices
• /
• v.24 no.3
• /
• pp.113-121
• /
• 2018

Position encoding is a method of applying weights according to position of words that appear in a sentence. Pointer networks is a deep learning model that outputs corresponding index with an input sequence. This model can be applied to coreference resolution using attribute. However, the pointer networks has a problem in that its performance is degraded when the length of input sequence is long. To solve this problem, we proposed two contributions to resolve the coreference. First, we applied position encoding and dynamic position encoding to pointer networks. Second, we stack deeply layers of encoder to make high-level abstraction. As results, the position encoding based stacked pointer networks model proposed in this paper had a CoNLL F1 performance of 71.78%, which was improved by 6.01% compared to vanilla pointer networks.

• Journal of KIISE:Software and Applications
• /
• v.29 no.11
• /
• pp.851-859
• /
• 2002

This paper discusses the design of a working system that visually recognizes hand gestures for the control of a window based user interface. We present a method for tracking the fingertip of the index finger using a single camera. Our method is based on CAMSHIFT algorithm and performs better than the CAMSHIFT algorithm in that it tracks well particular hand poses used in the system in complex backgrounds. We describe how the location of the fingertip is mapped to a location on the monitor, and how it Is both necessary and possible to smooth the path of the fingertip location using a physical model of a mouse pointer. Our method is able to track in real time, yet not absorb a major share of computational resources. The performance of our system shows a great promise that we will be able to use this methodology to control computers in near future.

• KIISE Transactions on Computing Practices
• /
• v.22 no.12
• /
• pp.625-631
• /
• 2016

Pointer Networks is a model which generates an output sequence with elements that correspond to an input sequence, based on the attention mechanism. A time complexity of the pointer networks is $O(N^2)$ resulting in longer decoding time of the model. This is because the model calculates attention for each input, if size of the input sequence is N. In this paper, we propose the pointer networks based on skip pointing model, which confirms the necessary input vector at decoding for reducing the decoding time of the pointer networks. Furthermore, experiments were conducted for the pronouns coreference resolution, which uses the method proposed in this paper. Our results show that the processing time per sentence was approximately 1.15 times faster, and the MUC F1 was 83.60%; this was approximately 2.17% improvement and a better performance than the original pointer networks.

• Journal of KIISE:Computer Systems and Theory
• /
• v.33 no.1_2
• /
• pp.35-43
• /
• 2006

An MPI Checkpointer is a tool which provides fault-tolerance through checkpointing The previous researches related to the MPI checkpointer have focused on automatic checkpointing and recovery capabilities, but they haven't considered portability issues. In this paper, we discuss design and implementation issues considered for portability when we developed an MPI checkpointer called STFT. In order to increase portability, firstly STFT supports the abstraction interface for a single process checkpointer. Secondly, STFT uses a user-based checkpointing method, and limits possible checkpointing places a user can make. Thirdly, STFT lets the MPI_Init create network connections to the other MPI processes in a fixed order. With these features, we expect STFT can be easily adaptable to various platforms and MPI implementations, and confirmed STFT is easily adaptable to LAM and MPICH/P4 with the prototype Implementation.

• Journal of KIISE:Software and Applications
• /
• v.32 no.7
• /
• pp.683-692
• /
• 2005

Java specification states that all accesses to an object must be checked at runtime if object refers to null. Since Java is an object-oriented language, object accesses are frequent enough to make null pointer checks affect the performance significantly. In order to reduce the performance degradation, there have been attempts to remove redundant null pointer checks. For example, in a Java environment where a just-in-time (JIT) compiler is used, the JIT compiler removes redundant null pointer check code via code analysis. This paper proposes a technique to remove additional null pointer check code that could not be removed by previous JIT compilation techniques, via early null pointer check using an architectural feature called predication. Generally, null point check code consists of two instructions: a compare and a branch. Our idea is moving the compare instruction that is usually located just before an use of an object, to the point right after the object is defined so that the total number of compare instructions is reduced. This results in reduction of dynamic and static compare instructions by 3.21$\%$ and 1.98$\%$. respectively, in SPECjvm98 bechmarks, compared to the code that has already been optimized by previous null pointer check elimination techniques. Its performance impact on an Itanium machine is an improvement of 0.32$\%$.

• Proceedings of the KIEE Conference
• /
• 2011.07a
• /
• pp.123-124
• /
• 2011

반도체 기술의 비약적 발전에 힘입어 현재의 개인용 컴퓨터는 고성능 CPU를 탑재하고 1990대의 텍스트 기반의 운영체제에서 벗어나 그래픽 기반의 운영체제에서 다양한 멀티미디어 기능을 제공 한다. 이를 위한 입력장치로 텍스트 기반 운영체제에서 주로 사용된 키보드뿐 아니라 마우스, 카메라, 터치스크린 등의 다양한 장치들이 사용되고 있다. 그러나 빔 프로젝터를 이용한 프레젠테이션의 경우 아직도 레이저포인터를 이용한 발표가 일반적이며 발표자와 빔 프로젝트용 PC와의 인터랙션이 없기 때문에 다양한 멀티미디어 기능 구현이 제한적이다. 본 논문에서는 USB 웹 카메라를 이용하여 프로젝터 화면을 촬영한 후 영상처리 라이브러리인 OpenCV를 기반으로 레이저 포인터의 위치와 동작을 검출하여 원거리에서도 사용자가 레이저 포인터를 이용하여 마우스 동작을 재현할 수 있는 시스템을 개발하고자 한다. 이를 활용하면 레이저 포인터를 사용하여 발표자가 별도의 입력장치 없이 PC와의 인터랙션이 가능해져서 다양한 멀티미디어 기반의 프레젠테이션이 가능해진다.