DC Field | Value | Language |
---|---|---|
dc.contributor.author | Kang, Haeyong | ko |
dc.contributor.author | CAO, NGUYEN NGOC | ko |
dc.contributor.author | Thang, Vu | ko |
dc.contributor.author | Yoo, Chang-Dong | ko |
dc.date.accessioned | 2019-11-28T05:21:02Z | - |
dc.date.available | 2019-11-28T05:21:02Z | - |
dc.date.created | 2019-11-27 | - |
dc.date.issued | 2019-11-22 | - |
dc.identifier.citation | 2019 대한전자공학회 추계학술대회 | - |
dc.identifier.uri | http://hdl.handle.net/10203/268670 | - |
dc.description.abstract | This paper considers a scene graph generation algorithm referred to as Relational Object Message Passing to enhance the representation of object relationships. In the scene graph, there are various kinds of relationship pattern depending on both subject and object. For example, given an image of a person and a bicycle, the relationship can be interpreted as the man riding, pushing, leaning against, or even falling off the bicycle. In this paper, for acquiring the good relational representation, two types of message passing algorithms are investigated: it operates on direct graph and undirect graph. The experiments showed the direct graph-based message passing is better than the undirect one, while comparing with the baseline. | - |
dc.language | English | - |
dc.publisher | 사단법인 대한전자공학회 | - |
dc.title | 장면 그래프 생성을 위한 물체간의 관계 표현 | - |
dc.type | Conference | - |
dc.type.rims | CONF | - |
dc.citation.publicationname | 2019 대한전자공학회 추계학술대회 | - |
dc.identifier.conferencecountry | KO | - |
dc.identifier.conferencelocation | 라카이 샌드파인 리조트, 강릉 | - |
dc.contributor.localauthor | Yoo, Chang-Dong | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.