• 한국측량학회지
• /
• 제39권6호
• /
• pp.599-607
• /
• 2021

스트리트뷰(Street-view) 영상은 도로의 특정 위치를 중심으로 한 전방위 영상을 제공하며, 보행 환경에 대한 다양한 장애물 정보를 포함한다. 보행자용 길안내 서비스에 활용하기 위한 보행 네트워크(Pedestrian network) 데이터는 교통약자를 비롯한 보행자의 이동 편의성을 보장하기 위하여 보행 장애물에 대한 최신 정보를 반영해야 한다. 본 연구에서는 스트리트뷰 영상과 딥러닝 기반의 객체탐지 알고리즘을 활용하여 서울 전역에 위치한 주요 보행 장애물인 볼라드(Bollard)를 학습하였다. 또한, 탐지된 볼라드 정보와 보행 네트워크 간의 공간매칭을 통해 횡단보도 노드를 대상으로 볼라드의 유무와 개수 정보를 장애물 속성으로 입력하고, 동시에 누락된 횡단보도 정보를 갱신하기 위한 프로세스를 정의하였다. 스트리트뷰 영상으로 학습된 모델은 보행 상황에서 스마트폰으로 촬영한 사진에 대해서도 적용이 가능하며, 향후 스트리트뷰 영상에 포함된 다양한 보행 장애물에 대한 추가 학습을 통해 효율적인 보행 장애 정보 갱신이 가능할 것으로 기대된다.

• ETRI Journal
• /
• 제31권4호
• /
• pp.429-437
• /
• 2009

This paper proposes a novel technique for 3D scene relighting with interactive viewpoint changes. The proposed technique is based on a deep framebuffer framework for fast relighting computation which adopts image-based techniques to provide arbitrary view-changing. In the preprocessing stage, the shading parameters required for the surface shaders, such as surface color, normal, depth, ambient/diffuse/specular coefficients, and roughness, are cached into multiple deep framebuffers generated by several caching cameras which are created in an automatic manner. When the user designs the lighting setup, the relighting renderer builds a map to connect a screen pixel for the current rendering camera to the corresponding deep framebuffer pixel and then computes illumination at each pixel with the cache values taken from the deep framebuffers. All the relighting computations except the deep framebuffer pre-computation are carried out at interactive rates by the GPU.

• 영미문화
• /
• 제14권2호
• /
• pp.1-21
• /
• 2014

Arne Naess' ecocentrally-oriented worldview of 'Self-realization' has been continually attacked by sociocultural critics since it was launched in the early 1970s. Especially ecofeminists and social ecologists criticize that the concept of Self-realization cannot accept social & cultural particularity enough. In other words, they assert that Deep Ecology cannot efficiently resist interpersonal hierarchy. Concentrating their criticism on compassion just for nonhuman beings, the interpersonal equality-oriented critics claim that Deep Ecology should incorporate voices of marginalized humans within their eco-centered world view. Even if Naess recently recognizes necessity to draw more attention to sociohistorical particularity, still unchanged remains essentialism in his neological term 'Self-realization.' Compared to exclusiveness in Naess' Self-realization, the Buddhist worldview of Interdependent Arising(pratityasamutpada) favors both intraspeciel egalitarianism and interpersonal equality as well. The very insight that all beings dependently co-originate reasons compassion for historically marginalized humans as well as nonhuman beings. Accordingly, today's environmentalism might better its efficiency as it goes toward the inclusive Buddhist world view of interdependent arising. For the human being tends to respond more readily to his or her individually urgent problems than their remote social matters.

• 대한임베디드공학회논문지
• /
• 제18권1호
• /
• pp.19-24
• /
• 2023

In traditional fish farming way, the workers have to observe all of the pools every time and every day to feed at the right timing. This method causes tremendous stress on workers and wastes time. To solve this problem, we implemented an automatic detection system for feeding time using deep neural network. The detection system consists of two steps: classification of the presence or absence of feed and checking DO (Dissolved Oxygen) of the pool. For the classification, the pretrained ResNet18 model and transfer learning with custom dataset are used. DO is obtained from the DO sensor in the pool through HTTP in real time. For better accuracy, the next step, checking DO proceeds when the result of the classification is absence of feed several times in a row. DO is checked if it is higher than a DO reference value that is set by the workers. These actions are performed automatically in the UI programs developed with LabVIEW.

• Journal of Multimedia Information System
• /
• 제8권2호
• /
• pp.85-92
• /
• 2021

In a face, there is much information of person's identity. Because of this property, various tasks such as expression recognition, identity recognition and deepfake have been actively conducted. Most of them use the exact frontal view of the given face. However, various directions of the face can be observed rather than the exact frontal image in real situation. The profile (side view) lacks information when comparing with the frontal view image. Therefore, if we can generate the frontal face from other directions, we can obtain more information on the given face. In this paper, we propose a combined style model based the conditional generative adversarial network (cGAN) for generating the frontal face from multi-view images that consist of characteristics that not only includes the style around the face (hair and beard) but also detailed areas (eye, nose, and mouth).

• Computers and Concrete
• /
• 제10권6호
• /
• pp.609-630
• /
• 2012

Currently, the design of reinforced concrete deep beams with web openings is carried out using empirical or semi-empirical methods and hence their scope of application is limited. In particular, high strength concrete deep beams with various web opening configurations have been given little treatment. In view of this, a nonlinear layered finite element method (LFEM) for cracking and failure analysis of reinforced concrete structures is used to conduct a parametric study to investigate reinforced concrete deep beams various web opening behaviours. This paper initially presents comparisons of LFEM output with published test results to numerical techniques. The paper then focuses on a parametric study on the shear strengths of deep beams with varying web opening configurations such as opening sizes and locations. The results confirm that the current design methods are inadequate in predicting the maximum shear strength when web openings are present. A series of parametric study offers insight into the maximum shear strength of the deep beams being critically influenced by the size and location of web openings.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2022년도 하계학술대회
• /
• pp.4-7
• /
• 2022

최근 다 시점 영상 콘텐츠 기반 3차원 공간(장면) 복원을 위한 다 시점 깊이 추정 딥러닝 네트워크 방법론이 널리 연구되고 있다. 다 시점 영상 콘텐츠는 촬영 구도, 촬영 환경 및 세팅에 따라 다양한 특성을 가지며, 고품질의 3차원 복원을 위해서는 이러한 특성을 이해하고, 적절한 깊이 추정 네트워크 기법들을 적용하는 것이 중요하다. 다 시점 영상 촬영 구도로는 수렴형, 발산형이 존재하며, 촬영 세팅에는 카메라 시점 간 물리적 거리인 baseline이 있다. 본 연구는 이와 같은 다 시점 영상 콘텐츠의 종류와 각 특징에 기반하여 콘텐츠(데이터 셋)의 특성에 따른 적절한 깊이 추정 네트워크 방법론을 다룬다. 실험 결과로부터, 기존의 다 시점 깊이 추정 네트워크를 발산형 또는 large baseline 특성을 가지는 데이터 셋에 곧바로 적용하는데 한계점이 존재함을 확인하였다. 따라서, 각 영상 환경에 적합한 '참조 시점 개수' 및 적절한 '참조 시점 선택 알고리즘'의 필요성을 검증하였다. 결론적으로, 3차원 공간(장면) 복원을 위한 딥러닝 기반 깊이 추정 네트워크 구현 시, 본 연구 결과가 다 시점 영상 콘텐츠 기반 깊이 추정 기법 선택에 있어 가이드라인으로 활용될 수 있음을 확인하였다.

• 한국정보통신학회논문지
• /
• 제24권11호
• /
• pp.1403-1410
• /
• 2020

최근, 비대면 경험 및 서비스에 관한 관심이 증가하면서 스마트폰이나 태블릿과 같은 모바일 기기를 이용하여 손쉽게 이용할 수 있는 웹 동영상 콘텐츠에 대한 수요가 급격히 증가하고 있다. 이와 같은 요구사항에 대응하기 위하여, 본 논문에서는 애니메이션이나 영화에 등장하는 명소를 방문하는 무대 탐방 경험을 제공할 수 있는 영상 콘텐츠를 보다 효율적으로 제작하기 위한 기법을 제안한다. 이를 위하여, Google Maps와 Google Street View API를 이용하여 무대탐방 지역에 해당하는 이미지를 수집하여 이미지 데이터셋을 구축하였다. 그 후, 딥러닝 기반의 style transfer 기술을 접목시켜 애니메이션의 독특한 화풍을 실사 이미지에 적용한 후 동영상화하기 위한 방법을 제시하였다. 마지막으로, 다양한 실험을 통해 제안하는 기법을 이용하여 보다 재미있고 흥미로운 형태의 무대탐방 영상 콘텐츠를 생성할 수 있음을 보였다.

• 대한수의학회지
• /
• 제2권1호
• /
• pp.37-50
• /
• 1962

Ten calves were subjected to general anesthesia with ten percent chlooral hydrate solution. The drug was administered by the method of slow intravenous injection so as to have a better control over the dosage and time until deep anesthesia was attained. Although one of ten calves failed to produce anesthesia, the remainder of nine responded satisfactorily with deep anesthesia. The dosage required averaged as great as 17.5gm per calf, and the time 23 minutes. In view of these advantages indicated in the results, further studies on the use of intravenous chloral hydrate for deep anesthesia in bovine species are to be justifiable.

• Journal of Information Processing Systems
• /
• 제17권3호
• /
• pp.505-517
• /
• 2021

This research presented an implementation of autostereoscopic virtual three-dimensional (3D) button device as non-contact style. The proposed device has several characteristics about visible feature, non-contact use and artificial intelligence (AI) engine. The device was designed to be contactless to prevent virus contamination and consists of 3D buttons in a virtual stereoscopic view. To specify the button pressed virtually by fingertip pointing, a simple deep learning network having two stages without convolution filters was designed. As confirmed in the experiment, if the input data composition is clearly designed, the deep learning network does not need to be configured so complexly. As the results of testing and evaluation by the certification institute, the proposed button device shows high reliability and stability.