http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
주 객체 위치 검출을 위한 Grad-CAM 기반의 딥러닝 네트워크
김선진,이종근,곽내정,류성필,안재형,Kim, Seon-Jin,Lee, Jong-Keun,Kwak, Nae-Jung,Ryu, Sung-Pil,Ahn, Jae-Hyeong 한국정보통신학회 2020 한국정보통신학회논문지 Vol.24 No.2
본 논문에서는 약한 지도학습을 통한 주 객체 위치 검출을 위한 최적의 딥러닝 네트워크 구조를 제안한다. 제안된 네트워크는 약한 지도학습을 통한 주 객체의 위치 검출 정확도를 향상시키기 위해 컨벌루션 블록을 추가하였다. 추가적인 딥러닝 네트워크는 VGG-16을 기반으로 합성곱 층을 더해주는 5가지 추가적인 블록으로 구성되며 객체의 실제 위치 정보가 필요하지 않는 약한 지도 학습의 방법으로 학습하였다. 또한 객체의 위치 검출에는 약한 지도학습의 방법 중, CAM에서 GAP이 필요하다는 단점을 보완한 Grad-CAM을 사용하였다. 제안한 네트워크는 CUB-200-2011 데이터 셋을 이용하여 성능을 테스트하였으며 Top-1 Localization Error를 산출하였을 때 50.13%의 결과를 얻을 수 있었다. 또한 제안한 네트워크는 기존의 방법보다 주 객체를 검출하는데 더 높은 정확도를 보인다. In this paper, we propose an optimal deep learning network architecture for main object location detection through weak supervised learning. The proposed network adds convolution blocks for improving the localization accuracy of the main object through weakly-supervised learning. The additional deep learning network consists of five additional blocks that add a composite product layer based on VGG-16. And the proposed network was trained by the method of weakly-supervised learning that does not require real location information for objects. In addition, Grad-CAM to compensate for the weakness of GAP in CAM, which is one of weak supervised learning methods, was used. The proposed network was tested through the CUB-200-2011 data set, we could obtain 50.13% in top-1 localization error. Also, the proposed network shows higher accuracy in detecting the main object than the existing method.
2차원 미디언 필터를 가진 레티넥스를 사용한 안개 구간에서의 차선 검출 방법
강지훈(Ji-Hun Kang),김창대(Chang-Dae Kim),류성필(Sung-Pil Ryu),김동우(Dong-Woo Kim),안재형(Jae-Hyeong Ahn) 한국정보기술학회 2015 한국정보기술학회논문지 Vol.13 No.10
This paper proposes new recognition method of road lanes foggy day. The method enables autonomous-driving of cars and the safety of the drivers while driving with bad visibility in foggy roads. The proposed method, firstly, determines whether the foggy or not from input images. If it is foggy then the Retinex algorithm with 2-D Median filter reduces foggy effects. And line segments painted on the road are detected by the advanced probabilistic Hough transform. Computer simulation results show that the proposed method could increase average line detection ratio about 2.7 times and recognize longer distance than earlier methods. Because the average processing time of the proposed method is about 5 frames per a second, it can use for real-time processing.
출력 화면에 적합한 H.264 비디오의 확대에 관한 연구
곽내정(Kwak Nae-Joung),권동진(Kwon Dong-Jin),류성필(Ryu Sung-Pil) 한국콘텐츠학회 2006 한국콘텐츠학회 종합학술대회 논문집 Vol.4 No.2
최근 멀티미디어 데이터의 발달로 다양한 유무선 통신환경에서 멀티미디어 데이터가 서비스되고 있다. 이러한 멀티미디어 데이터는 이동전화, PDA, 노트북, PC 등 다양한 화면의 크기를 가지는 디스플레이 장치로 디스플레이된다. 따라서 사용자의 디스플레이 장치에 따라 원영상의 크기를 조절할 필요성이 대두된다. 본 논문은 H.264 비디오를 사용자 화면에 적합하도록 크기를 변경하는 방법을 제안한다. 기존의 연구는 대부분이 영상의 크기를 줄이는 것이 대부분이었으나 본 논문에서는 해상도를 증가하여 서비스하는 방법에 관해 제안한다. 제안방법을 CIF 및 QCIF 영상에 적용한 결과는 영상이 블록킹 현상 확대됨을 보여준다. Lately, development of multimedia industry has multimedia data serviced through various wire?wireless communicating environment. These multimedia data is displayed on the device of various sizes such as mobile telephone, PDA, notebook, PC. Therefore, there has needed to resize multimedia data appropriate to the screen size of user's display device. In this paper, we propose the method to resizing H.264 videos sequence fitting to the screen size of user's display device. The existing most methods reduce the size of video sequence, but this paper proposes the upsampling method. The results show the proposed method upsamples video sequences without blocking effects.
김창대(Chang-Dae Kim),최서혁(Seo-Hyuk Choi),류성필(Sung-Pil Ryu),김동우(Dong-Woo Kim),안재형(Jae-Hyeong Ahn) 한국정보기술학회 2016 한국정보기술학회논문지 Vol.14 No.10
The flickering phenomenon of LED lamps may reduce the detection rate of the camera positions significantly. In order to address this problem, we propose the method that detects positions of smartphone cameras in order to put security stickers on it automatically. The method can reduce the security checking time that a guard attaches the stickers or checks whether they have been damaged or not at rush hour. The proposed method uses the adaptive frequency degree, applying appropriate weights according to predictive location and size of camera position after binarization with the adaptive threshold. The computer simulation results show that the proposed method increases the detection ratio than the previous ones in the LED lighting environment.
연산호 육성용 어초 개발을 위한 송이 다공성 콘크리트의 적용성에 관한 실험적 연구
홍종현(CHONG-HYUN HONG),김문훈(MOON-HOON KIM),김석철(SEOK-CHEL KIM),박성배(SUNG-BAE PARK),류성필(SEONG-PIL Ryu) 한국해양공학회 2005 한국해양공학회지 Vol.19 No.4
In the study application of volcanic scoria concrete to artificial reefs is investigated. Volcanic scoria is a natural volcanic product that shows light weight, milli-porous, and far-infrared irradiation characteristics. The properties of volcanic scoria concrete using Jeju scoria aggregate are evaluated by conducting a comprehensive series of tests on strength and void ratio. It is concluded that the volcanic scoria concrete has the sufficient strength of 4MPa-13MPa and adequate void ratio of 12%-35% to be accepted as artificial reef concrete. The field experiments are performed through observation by scuba diver's at the Seogwipo coast. Porous specimen and plane concrete specimen are prepared for comparison purposes. Seasonal changes of soft coral on the two series of test specimens were have been observed from Apr. 9, 2004 to Mar. 18, 2005. The soft coral is well grown on the porous specimen however there are no significant changes on the conventional plain concrete specimen. Thus it is concluded that the volcanic scoria concrete is highly suitable as artificial reef concrete.