DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Han, Dong Su | - |
dc.contributor.advisor | 한동수 | - |
dc.contributor.author | Lee, Byeok San | - |
dc.date.accessioned | 2018-06-20T06:23:19Z | - |
dc.date.available | 2018-06-20T06:23:19Z | - |
dc.date.issued | 2017 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=718713&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/243385 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 전기및전자공학부, 2017.8,[iii, 23 p. :] | - |
dc.description.abstract | Closed circuit televisions (CCTVs) have been installed in many public places for its own purpose. For example, CCTV operates to prevent crimes or to manage traffics. For this and many reasons the number of CCTVs has been increasing and will increase much faster with the growth of IP cameras and Internet of Things. For these cameras to be used efficiently it is necessary for CCTVs to detect various events automatically. However, it is general to detect pre-defined events because learning detection for a specific event takes high cost. This paper addresses a method of deep learning for automatic event detection learning with web data. Deep learning requires data, so they are collected from the web and used as training data. To search the images, search keywords are automatically generated utilizing WordNet which is a lexical database for English. With those keywords, images are retrieved from the web search results. Also, we applied unsupervised learning to downloaded image data which are noisy and applied generalization technique to learn data which are not event data. It is shown that these approaches are effective for event detection. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Deep Learning▼aEvent Detection▼aUnsupervised Learning▼aGeneralization | - |
dc.subject | 딥 러닝▼a사건 감지▼a비지도 학습▼a일반화 | - |
dc.title | Deep learning for event detection using web data | - |
dc.title.alternative | 웹 데이터를 활용한 사건 감지를 위한 딥 러닝 방법 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :전기및전자공학부, | - |
dc.contributor.alternativeauthor | 이벽산 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.